This is an insurance coverage case of great financial magnitude and complexity. Both sides appeal legal decisions made during a long trial, aspects of the jury verdict, and a final judgment and judgment notwithstanding the verdict entered and certified for appeal by the trial court. We AFFIRM some decisions of the trial court, REVERSE others, and REMAND, primarily to allow the parties to show how much of the alleged damages are covered by the insurance policies.

The insured parties, the plaintiffs-appellees/cross-appellants, are Enserch Corporation ("Enserch"), a Texas-based engineering firm, and its New York-based subsidiary, Ebasco Services, Inc. ("Ebasco").1 In April 1982 the two companies jointly took out two "claims made, prior acts" policies (covering any claims, including those for acts committed before the policy period, made against the insured during the policy period for "any act, error or omission" in its performance of professional services). One policy was with General Accident Insurance Co. ("GA"), the other was with Evanston Insurance Co. ("Evanston") (collectively, "the insurers"), and each had a maximum recovery per claim of $25 million. The Evanston policy covered "lawyers, accountants, management consultants, risk management consultants, business and economic research consultants, corporate training consultants, and tax consultants". The GA policy covered architectural, engineering, and construction services. The policies had individual deductibles for each claim of $5 million, an amount reduced to $500,000 after the insured had paid a sum of $7.5 million (the "aggregate deductible") for claims made during any policy period. The deductible endorsement of each policy stated that any deductible would apply toward the aggregate deductible of both policies. For the Evanston policy the annual premium was $25,962.50; for the GA policy it was $1,075,000.

The application for insurance asked the insured to state if it knew of any circumstances that might give rise to claims against Ebasco. Ebasco's attorneys had thirty-six of its officers respond to a polling; none of their responses referred to the Washington Public Power Supply System ("WPPSS") project that gave rise, in February 1983, to the lawsuit underlying this case.

In addition to that specific polling, several provisions of each policy either apply to Ebasco's knowledge or representations or may exclude the liability for which it now seeks coverage.

Section I(c) of the GA policy specifically precludes coverage of claims for "any act, error, or omission of which any director, partner, or officer of the company had any knowledge at the effective date of the policy". Under condition VIII of the GA policy the insured agrees that all statements in the application are "agreements, representations and warranties". The GA policy also specifically excludes coverage for "estimates of probable construction cost or cost estimates being exceeded" and for "advising or requiring, or failure to advise or require or failure to maintain or procure any financing for any portion of any project or of services or labor connected with such project". The insurers contend that these exclusions apply to any cost estimates or financial advice regarding the WPPSS project.

Condition 1 of the Evanston policy provides that all statements made in the application are "personal representations ... and that this policy is issued in reliance upon the truth of such representations...." The Evanston policy also specifically excludes claims based on "dishonest, deliberately fraudulent, malicious or knowingly wrongful acts, errors or omissions".

Ebasco was the architect-engineer for WPPSS Project 5, the fifth nuclear power generating station in one of the country's largest nuclear construction projects. The bonds that financed WPPSS Projects 4 and 5, unlike those for Projects 1-3, were not backed by the federal government. Rather, 88 cities and utilities in the Northwest guaranteed them through take-or-pay obligations. Because of excessive costs, construction on Project 5 stopped in May 1981 (when the project was only 14% completed), and was terminated permanently in January 1982 (three months before the effective date of the two policies in question). In February 1983 the holders of WPPSS bonds backing Projects 4 and 5 (infamous as the "Whoops" bonds) filed a class action lawsuit (the Multidistrict litigation ["MDL"] lawsuit) against over 100 defendants (including Ebasco) for the $2.25 billion of bonds that had already been issued to fund the two projects.

In connection with its work on Projects 3 and 5, Ebasco had been required to submit a series of letters providing cost estimates and status reports for its work on each project.2 These letters were used to inform underwriters and public bond purchasers of progress on the plant. In fourteen such letters issued between February 1977 and March 1981, Ebasco described the finances and progress of Project 5 in terms that were (for whatever reason) untimely and, at least, overly optimistic.3 The insurers and the bondholders who sued Ebasco (for violating federal and state securities laws, for common law fraud, negligence, and professional malpractice) argued that Ebasco deliberately misrepresented the costs of Project 5 to sell more bonds and to keep the project going. After many of the utility guarantors of the bonds obtained declarations absolving them of contractual obligations to back the bonds, defendants like Ebasco were the only parties remaining to foot the large bill for the failed project.

Although the insurers contend that Ebasco did not originally expect them to provide its defense, the insurers did inform Ebasco that it would issue a reservation of rights letter in providing a defense; no such letter was ever written. In October 1984 Ebasco finally did demand coverage and a defense for the claims in the MDL suit. On May 15, 1985, the insurers, citing Ebasco's non-cooperation, filed a declaratory judgment action in New York to determine their duties under the policies. Ebasco achieved a stay of that action after filing a state court action in Texas (removed by the insurers to federal court) for coverage under the policies.

In February 1989 Ebasco was one of the final defendants to settle claims from the MDL bondholder suit. The two-tier settlement was for $50 million, payable as follows: $7,166,666.67 in cash from Ebasco and $42,833,333.33 from the two insurance companies, which Ebasco obligated itself to sue. The agreement further provides that Ebasco will receive 25% of any recovery on the GA policy and 50% on the Evanston policy until it has recovered the amount of its cash payment, its costs to defend the MDL suit, and its costs to bring this action; Ebasco will keep any costs awarded to it separately in this suit. If the bondholders do not receive at least $3 million from this suit, then Ebasco promised to pay that additional amount in cash. Finally, the MDL plaintiffs and Ebasco agreed to split any costs recovered in its forthcoming action for insurer bad faith. The trial court found this settlement to be reasonable, and approved it under Fed.R.Civ.P. 23(e).

After a 45-day trial on the insured's demand for indemnification of its liability to the MDL plaintiffs,4 the jury reached a verdict (entered as a final judgment on May 18, 1990) in favor of Ebasco, granting it the full $50 million of coverage under both policies; attorney fees of $6,381,186 for defending the MDL suit, $4,190,204 for the trial in this case, $140,000 for pursuing successful post-trial proceedings, and $200,000 for defending a successful appeal to this Court; in-house defense costs incurred in the MDL lawsuit of $2,652,693; and other costs of $1,211,775 incurred in defending the MDL suit.

Judge Robert B. Maloney, the trial judge, later granted the insurers' request for a JNOV, and reduced the verdict in the following ways. First, he released from liability in this stage of the case Shand Morahan, underwriter and managing agent for the policies from liability in this stage of the case.5 Second, he rejected the jury's finding of no mutual mistake and decided that the defendants had proved by clear and convincing evidence that both parties intended to coordinate the two policies (with a maximum recovery of $25 million) if a single claim involved both policies. Third, he decided that a deductible of $7.5 million should be applied against the jury's damages award; he applied it specifically to the outside attorney fees and expenses incurred by Ebasco in defending the MDL lawsuit. Therefore the total of those costs--$7,592,961--was reduced to $92,961. Fourth, he reversed the award of $2,652,693 for in-house legal expenses. Even though the jury found such costs to be reasonable, the court found that the plaintiffs had failed to show that the costs were direct or consequential damages from the insurers' failure to defend. Fifth, and finally, the court denied the plaintiffs their attorney fees (granted by the jury) for work in connection with the trial below, the post-trial motions, and this appeal, because they (who first sought the application of Texas law in this case) did not make a proper demand for such fees under Tex.Civ.Prac. & Rem.Code § 38.001 or § 38.002(2) and (3). Ebasco appeals the second, fourth, and fifth of these decisions.6

On the same day (August 6, 1990) the trial court issued an amended final judgment declaring GA and Evanston jointly and severally liable for $25 million under the two policies, plus the $92,961 in outside attorney fees, with prejudgment interest at 6% from May 15, 1985 (when the insurers filed their declaratory judgment action in New York) until May 18, 1990, and a post-judgment rate of 8.70% until judgment is paid. In a decision after the certification for appeal of its amended final judgment and the JNOV, the trial court denied the plaintiffs' bill of costs, except for Special Master fees of $101,081. Both sides appeal aspects of the final judgment.

The insurers appeal decisions made by the trial judge in deciding certain matters of law against it, for erring in his interrogatories to the jury, and for denying all of its possible defenses against Ebasco for fraud and misrepresentation in obtaining the insurance policies. Ebasco, generally pleased with all decisions made at the trial, is appealing the judge's various decisions reducing their judgment from more than $65 million (excluding interest) to just over $25 million. The insurers want a chance to retry the case under more favorable legal rules; the insured wants to restore the victory it was deprived of by the trial court's JNOV.

We discuss first the decisions made at trial appealed by the insurers, and then reach the decisions in the JNOV and final judgment appealed by both sides.

In a pretrial order the trial court held that Evanston had a duty to defend Ebasco in the MDL litigation; the court decided sua sponte that GA had the same duty. The jury found that both Evanston and GA had breached those duties. The insurers challenge the trial court's ruling.

This Court reviews de novo a decision of law such as the determination of a duty to defend, which depends in this case on the language of the MDL plaintiffs' complaint and the terms of the applicable insurance policies. If any allegation in the complaint is even potentially covered by the policy then the insurer has a duty to defend its insured.7 We reach the same result as the trial court, which applied the "complaint/allegation" rule with appropriate liberality here.8

1. The Evanston Policy. The insurers contend that the MDL complaint did not implicate the Evanston policy because it only sued Ebasco as an engineer and construction manager of the WPPSS projects; these professions are covered by the GA policy. The MDL complaint, however, did clearly implicate professions that were covered by the Evanston policy: lawyers, accountants, and economic and management research consultants. The insurers contradict themselves by arguing, as we note below, that the GA policy did not apply because the complaint referred to cost estimates; they cannot so argue while also denying that the professions covered by the Evanston policy undeniably do involve cost estimates. Evanston had a duty to defend Ebasco in the MDL suit.

2. The GA policy. GA contends that specific exclusions in its policy applied to the MDL complaint and excused its refusal to defend. We disagree. Whether or not the exclusions for "cost estimates" and financial advice and procurement were intended to apply to situations such as the MDL litigation,9 the MDL complaint did refer to inaccurate reports regarding design, engineering, and construction. All of these activities were covered by the GA policy. The trial judge would have misapplied the "complaint/allegation" rule under Texas law by holding that all of the activities alleged in the MDL complaint fell outside the GA policy.

B. The Duty to Indemnify and Allocation.

Both insurers had a duty to defend Ebasco in the MDL suit. Both insurers breached that duty.10 An insurer that breaches its duty to defend, however, does not necessarily owe its insured complete indemnification for a settlement the insured reached on its own. It is axiomatic insurance law that the duty to defend is broader than the duty to indemnify or pay. Some cases have suggested that by breaching its duty to defend, however, an insurer is estopped from challenging its own liability for the insured's settlement of the case it refused to defend.11 Courts applying Texas law have made such a broad suggestion, however, only in cases where the settlement or judgment below actually addressed whether the liability would be covered by insurance.12 In this case, that issue has not yet been addressed.

Although the insurers are bound by the settlement Ebasco arranged for itself in this case, they are not estopped from contesting coverage of that liability. Although the insurers owed Ebasco a duty to defend, they do not, even by having breached that duty, necessarily owe it a duty to pay. This Court very recently affirmed that a "finding of [the insured's] liability, to which [the insurer who breached its duty to defend] is now bound, is distinct from the question of coverage, which cannot be created ex nihilo by estoppel".13

The jury in this case was instructed as follows:

To determine whether the exclusions preclude coverage under the policies you must consider only the allegations [against Ebasco] contained in the bondholder Complaints. No other evidence is relevant for deciding whether Plaintiffs have met their burden with regard to the policy exclusions.

The jury was thus allowed to conflate the insurers' duty to indemnify with their duty to defend. The trial court, in labeling the jury's decision on indemnification an "all-or-nothing proposition", denied any apportionment of that indemnification between damages covered and those not covered by the policies. This case must be remanded for findings to make the necessary apportionment between damages for which the insurers owe, and those for which they do not owe, Ebasco a duty to pay.

We appreciate the difficulty of this apportionment decision, and we are reluctant to force any insured to retry a case that it settled after its insurer unjustifiably denied its defense; an insurer should not benefit from breaching its duty to defend. Nevertheless, in such a situation, we cannot allow an insured to settle allegations against it (some of which might be covered by its insurance, some of which might not) for its policy limits and then seek full indemnification from its insurer when some of that settled liability may be for acts clearly excluded by that policy.

There are several possible sources for help in allocating the claimant's damages. There are the allegations contained in the bondholders' complaint; there is the settlement between Ebasco and the bondholders; and there are the facts that would have been the subject of the MDL lawsuit, had it been tried.

We can imagine cases where the allegations alone could sufficiently justify an allocation of damages. In such a case the judge could compare the insurance contract with the complaint and determine as a matter of law what portion of the damages were covered. This is not such a case. The MDL allegations are so insufficient to decide the allocation issue, in fact, that both parties before us have argued that the bondholders' complaint justifies an allocation entirely to their benefit as a matter of law. Neither is right. On remand the trial court will have to look further.14

We consider it unlikely that the settlement in this case (perhaps in any case) will be sufficient to justify an allocation of damages. A settling insured will never include a statement that would admit liability in a way that deprived it of indemnification, and a settling insured that allocated all of the damages to covered claims would be too susceptible a target of an insurer arguing (as the insurers argue here) that the settlement was presumptively collusive. For this reason, and because Texas law forbids insurers who have breached their duty to defend from challenging the reasonability of the insured's settlement (we discuss this below), we think that the settlement in this case will be of little help. The final answer on this, however, is for the trial court to decide. If a trial court could apportion damages solely by applying the terms of the settlement, then the decision would be one of law, not requiring a jury.

If the apportionment cannot be made as a matter of legal interpretation of either the allegations in the complaint or the settlement agreement itself, then the trial court will have to call a jury to allocate damages through the most limited trial it considers necessary. Unfortunately, no cases cited by either party, nor any case we can find, gives us or the trial court specific guidance on how to proceed. The jury can consider any facts that could have been considered in the MDL lawsuit itself. The trial court, of course, will have considerable leeway in deciding what facts are truly relevant to the apportionment decision, and can thereby (we hope) prevent a full-blown retrial of the MDL lawsuit. The jury will not be considering whether Ebasco was liable to the bondholders for any loss greater than the $50 million settlement, or what the damages award might have been after the MDL trial. It will be considering only what portion of the $50 million liability to which Ebasco did subject itself and its insurers was covered by the two insurance policies in question.

Because the findings for which we remand this case would have been made had the insurers fulfilled their duties to defend Ebasco,15 and because we do not want to encourage insurers to breach those duties, we also rule that the insurers will pay Ebasco its reasonable costs and attorney fees for defending its side of the allocation determination on remand.

We note at least one considerable limitation on evidence the insurers can either seek through discovery or use in apportioning the damages. Texas law denies insurers like these a collateral attack on the settlement itself. The insurers challenge the trial judge's denial of discovery into the reasonableness of Ebasco's settlement, and his refusal to submit this issue to the jury. Recent opinions of both this Court and the Texas Supreme Court have confirmed that, unlike a request for allocation, an attempt to contest the reasonableness of a consent judgment entered into between the insured and an injured third party is unavailable to an insurer who has wrongfully breached its duty to defend.16 The trial judge wisely prevented such a collateral attack by the insurers below; in refusing discovery into that settlement he relied on the sufficiency of the reasonableness hearing by the judge in the MDL lawsuit. Unable to find any signs of collusion or fraud in the two-tier settlement between Ebasco and the bondholders, the insurers can suggest only that we adopt a presumption of fraud underlying such settlements. We reject that suggestion and affirm both the longstanding Texas rule and the trial judge's decision in this case. In defending its arguments for allocation the insurers cannot attack the reasonability of the settlement itself, or of the manner in which it was reached.17

The insurers assert three additional defenses denied or decided against them by the trial judge. The first is the trial court's decision to allow the jury to apply a subjective standard to the insurers' common law misrepresentation defense. The second is the denial of their defenses based on what they contend are warranties in Ebasco's applications for insurance. The third is the trial court's refusal to submit a defense for GA under Section I(c) of its policy, as to which standards and burdens of proof different from those associated with common law misrepresentation might apply. Only the last issue requires reconsideration on remand.

1. The insurers contend that a subjective standard is inappropriate for evaluating Ebasco's alleged misrepresentations in response to the polling. They would have this Court apply an objective ("knew or should have known") standard to their defenses under common law misrepresentation. Their only support for this standard comes from other jurisdictions. A misrepresentation defense under Texas law requires a showing that "the misrepresentation was made willfully with the intent to deceive or to induce the insurance company to issue the policy".18 An applicant for insurance cannot willfully intend to deceive its potential insurer unless it actually, not constructively, knows that what it misrepresents is untrue, especially when the allegedly untrue statement was one of opinion. Further, Texas law holds that even a material misrepresentation in an insurance application does not defeat recovery if it is made innocently and in good faith.19 A good faith standard implies subjectivity; an objective standard could not fit within a test that allowed innocent misrepresentations.

In explaining the defense of misrepresentation, the trial court instructed the jury to decide "whether any person who responded to the polling ... believed at the time of making the polling representation, that those facts might give rise to a future claim". This instruction, as well as the specific interrogatory answered by the jury,20 applied Texas law correctly. As to this most aggressively contested subject of the trial, the jury apparently chose to believe Ebasco's witnesses. In spite of what the insurers call "ample evidence" contradicting those witnesses, there was evidence to show that the officers of Ebasco did not expect to be sued for the failure of the WPPSS project. We affirm the trial court's instructions, and we refuse to reject the jury's finding, on the insurers' common law misrepresentation defense.

2. The insurers also contend that Ebasco warranted the truth of its responses to the polling. Unlike the defense of common law misrepresentation, a defense based on breach of warranty might place on Ebasco the burden of proving that its officers knew their responses to the polling to be false, and might judge that knowledge by an objective standard. The insurers therefore challenge the trial court's refusal to allow them a separate defense based on these alleged warranties, even though the defense arises from the same facts on which the common law misrepresentation defense is based. We will consider the alleged warranties in each policy.

The Evanston policy contains no warranty. It states that "the statements in the application are personal representations," and it makes no mention at all of warranties.

The GA policy is more complicated. It states that "all Insureds agree ... that the statements in the application are their agreements, representations, and warranties " (emphasis added). Ebasco argues that by combining these three terms the policy created an ambiguity that must be interpreted in favor of the insured. Although some statements in the application may be thereby warranted, it is impossible to determine whether any individual statement is to be considered an agreement, a representation, or a warranty. Further, the word "REPRESENTATION" appears above the signature line of the GA application.21 GA contends that this ambiguity is only a sign of lawyerly redundancy, and that we should consider every statement in the application a warranty, as well as an agreement, and a representation. We disagree. The different effects of representations and warranties are well known. For the purpose of insurance law, an applicant who merely represents some fact does not warrant it. Texas law obligates us to interpret ambiguities in insurance applications against the insurer; in this case common sense reaches the same result. To subject its potential insureds to the legal effects of a warranty, an insurer should make the terms of that warranty unambiguous and clear. In this case, it did so in a manner that is so unclear as to be unenforceable by this Court.

3. The insurer's claimed defense under Section I(c) of the GA policy is more problematic.22 GA contends that the trial court erred by not allowing it to raise Section I(c) as a separate defense; GA suggests that, unlike the insurers' joint defense of common law misrepresentation, Ebasco would have borne the burden of proving that coverage of the bondholder claims was not excluded by Section I(c).

Ebasco argues that a defense under Section I(c) would not differ from the defense under common law misrepresentation, which the jury decided in Ebasco's favor by finding that Ebasco did not make false representations in response to the polling. Section I(c), however, is not limited to statements made in application for insurance. Nor is it limited to the insured's belief about claims that might arise against it. Section I of the GA policy defines coverage for a "claims made" policy. Subsection (c) reads: "No director, partner, or officer of vice presidential level or higher of the Insured has knowledge of such act, error or omission on the effective date of this policy." GA interprets this clause to exclude coverage for acts, errors, and omissions "which [any] officer knew or reasonably should have known might give rise to a claim under the policy". This interpretation misreads what we consider the section's plain meaning in two important ways: it defines it to apply not to any "act, error or omission", but to acts, errors, or omissions that might give rise to a claim against the insured; and it applies a subjective standard to the insured's knowledge. The first change makes Section I(c) duplicate what Question No. 13 already presented to the jury; the second change allows the court to judge the answers Ebasco's officers made in the polling according to a different standard.

We reject GA's characterization of Section I(c) as a warranty. In a similar case arising under Louisiana law (where that distinction would make no difference), Judge Alvin Rubin of this Court interpreted a coverage clause in a "claims made" policy addressing the insured's knowledge of "any circumstance" that might result in a claim being made against it. Judge Rubin rejected an argument made by the insurer that, because similar questions were answered in the original application for insurance, the coverage clause should be applied against the insured as a warranty. He wrote that "[a]lthough the information on the application may have served as a warranty or representation ..., the specific policy provision in question is patently a limitation on coverage.... [T]he issue before us relates only to the scope of the insurance [the policy] affords."23 We apply the same reasoning to this case. Section I(c) of the GA policy contains one of the conditions for coverage which an insured must satisfy, and which Ebasco has the burden of proving it satisfied in this case, before it is entitled to indemnification by its insurer.24

The trial judge correctly instructed the jury that Ebasco bore the burden of proving that it had satisfied the conditions for coverage set forth in the policies, yet Question No. 1 presented to the jury askedonly: "Do you find by a preponderance of the evidence that one or more of the claims made against Ebasco in the bondholder litigation arose out of the professional services covered by the General Accident policy?" (emphasis added). Again, the trial court erred in letting the jury confuse the duty to defend with the duty to pay. Ebasco bears the burden of proving not only that one of the bondholder claims was covered by the GA policy, but that all the claims for which it seeks indemnification were covered by the policy. Because the proper application of the coverage provisions of the GA policy will be accomplished when the trial court allocates Ebasco's liability between covered and uncovered claims, it will not be necessary to force an entirely new trial for the proper consideration of this issue. We accept GA's argument only insofar as it places the burden on Ebasco to show that "[n]o director, partner, officer of vice presidential level or higher of the Insured has knowledge of such act, error or omission" for which it now seeks coverage under the GA policy. This should not, as Ebasco warns us, place on it an intolerable burden of proving a negative. Rather, it places on Ebasco the burden felt by any insured who must make the preliminary showing that the claim for which it seeks coverage is indeed covered by its policy with the insurer. The responses to the polling should go far, and perhaps be sufficient, to answer the fact question of whether or not Ebasco's officers actually did know of acts for which the MDL plaintiffs sued, a fact that matters only insofar as the court or jury below finds that the settlement between Ebasco and the bondholders was intended to compensate the latter for such acts.

A. Prejudgment Interest.

The trial court applied a 6% rate of prejudgment interest under its interpretation of Texas law.25 That interpretation was incorrect. This Court has confirmed recently that the 6% prejudgment interest rate of article 5069-1.03 applies only "in cases in which the court, having already found liability, concluded that the contract or account unambiguously established the amount owed".26 Insurance contracts as thoroughly disputed as those before us do not unambiguously establish the amount owed. The insurers themselves, in denying that Ebasco is legally liable for a sum certain, argue that the MDL settlement (to say nothing of the insurance contracts themselves) did not establish an unambiguous amount owed.

When article 5069-1.03 is inapplicable, as here, a court should award prejudgment interest according to the equitable principles of article 5069-1.05. Article 5069-1.05 directs a court to apply prejudgment interest at the rate of 10% to prevailing parties in all but the most exceptional circumstances.27 That rate applies here.

B. Accrual of Prejudgment Interest.

Because article 5069-1.05 applies prejudgment interest according to equitable principles, the rule that such interest under article 5069-1.03 should be applied from the date on which the insurer contests coverage may not apply. The district court ruled that prejudgment interest should accrue from the date the insurers filed their declaratory judgment action against Ebasco (May 15, 1985). Ebasco argues that accrual for such a long period is justified because the party ultimately being compensated is the bondholder class of the MDL suit. The bondholders are not a party to this action, however, and their recovery is not the subject of this case.

Although recent Texas cases have applied prejudgment interest more broadly, they have also limited its purpose to the recovery of accrued damages, and insisted that it not be imposed for a punitive effect.28 To award prejudgment interest against the insurers for four years before Ebasco was actually liable to the bondholders for any sum would be wrong. Prejudgment interest should not begin accruing until the insurers' breach caused Ebasco an actual loss: that is, when Ebasco became liable to the MDL plaintiffs on the date of their settlement (February 28, 1989). And such interest should only apply to the sum for which Ebasco actually is liable: the $7,166,666.67 that constituted the first "tier" of their settlement agreement. To award prejudgment interest on a larger amount, or to let it accrue for a longer period of time, would not correspond with the limited equitable purposes of article 5069-1.05 as Texas courts have interpreted that law. The trial court erred in ruling otherwise. Prejudgment interest at the rate of 10% should apply to the $7,166,666.67 for which Ebasco itself was liable, and it should begin accruing on February 28, 1989.29

C. Prejudgment Interest on Attorney Fees.

The trial court correctly refused to apply prejudgment interest to any award of attorney fees to Ebasco. The Texas Supreme Court has clearly precluded such an award when attorney fees are awarded under article 5069-1.03.30 This Court has cited Hervey for the proposition that under Texas law prejudgment interest is never allowed on awards of attorney fees.31 A panel of this Court has specifically extended Hervey to awards of attorney fees under Tex.Civ.Prac. and Rem.Code § 38.001 (the statute applicable to this case).32 Prejudgment interest will not apply to any award of attorney fees in this case.

D. Attorney Fees.

Ebasco sought recovery of $4,530,204 in attorney fees for defending itself in this lawsuit. The jury found that amount to have been reasonable and necessary to Ebasco's defense. The trial court, however, refused to award it because of Ebasco's failure to cite the pertinent Texas statute in its amended complaint.33 The trial court felt, with some justification, that Ebasco was pursuing the strict application of Texas law in every area except the one where it worked to its disadvantage. We understand that frustration and we acknowledge that Ebasco never formally cited the statute it asks us to enforce; nevertheless, holding an insured in Ebasco's shoes to such strict pleading standards is unsupported by case law and was incorrect in this case.

Our Court has suggested that "a party must plead entitlement to [§ 38.001] fees at least with some particularity".34 Yet in general courts appear to interpret the "presentment" requirements of § 38.002(2) (which obligates the claimant under § 38.001 to "present the claim to the opposing party") more liberally.35 Some Texas courts have even suggested that such an award is mandatory where an attorney prevails on a contract claim and satisfies the requirements of § 38.002.36 We can find no case where a party that clearly presented a claim for attorney fees for enforcing a contract (the situation § 38.001(8) addresses), yet failed to mention that statute, has been denied those fees. We find that the insurers were clearly put on notice that Ebasco, if it prevailed, would seek its attorney fees under the applicable statute, and that to deny Ebasco those fees would hold them to too rigid a pleading requirement. Ebasco has prevailed on its defense claim. The insurers therefore owe Ebasco $4,530,204 in attorneys' fees incurred for prevailing on that portion of its contract claim. The insurers will be liable to Ebasco for additional attorneys' fees should Ebasco prevail on its claim for indemnification.

E. In-House Costs.

The trial court's JNOV denied Ebasco recovery for its in-house defense costs. Ebasco contends that the trial court erred in finding that Ebasco "would have incurred all or substantially all of the expenses incurred whether or not Defendants originally offered to defend that MDL 551 litigation" and that Ebasco had "failed to properly segregate" in-house costs between those caused and those not caused by the insurers' breach of the duty to defend. The jury found only that the sum requested by Ebasco, $2,652,693, was "reasonable and necessary". It did not find that Ebasco would have spent the same amount of money if the insurers had defended Ebasco in the MDL litigation. The trial court's finding of fact on this issue, therefore, is only reversible if it is clearly erroneous. Although the trial court could have supported this finding more fully, we cannot say that it was clearly erroneous. Although the insurers presented no evidence to show that Ebasco's in-house costs were not caused by its breach, Ebasco itself presented only one witness to defend these costs. He stated that "it was plainly more cost efficient for Ebasco to divert some of its own people ... than to go outside and have to hire engineering consultants or lawyers to get involved ... and do it at a higher price". The trial court presumably considered this broad statement speculation, unsupported by evidence sufficient to support the award Ebasco sought. As a matter of fact, this decision was not clearly wrong. Almost all insureds incur costs related to their defenses--costs for which their insurers do not reimburse them. If Ebasco sought to distinguish some of those costs from those that it would have incurred even if the insurers had provided its defense, it failed to provide enough evidence to let the trial court make the necessary distinction.

Nor can we say that the trial court applied the law incorrectly. Ebasco cites us to no case from this jurisdiction where insureds were indemnified for all in-house costs when an insurer breached its duty to defend.37 We are unprepared (especially in a case where Ebasco did hire very competent, and expensive, outside counsel) to establish a general presumption favoring the recovery of in-house costs by abandoned insureds. We do not think insurers should ever gain from improperly refusing to defend their insureds. But such a rule, if it awarded an insured with recovery for in-house costs that would have been incurred regardless of who supplied the defense, could also improperly profit an insured. The insured must show that its additional costs were caused by the insurer's breach.

F. Bill of Costs.

Ebasco also appeals the trial court's denial of many of its court costs incurred in trying Count One of this action. It did not seek these costs, and the trial court did not award (and deny) them, until after the JNOV and Amended Final Judgment had been certified for appeal pursuant to Fed.R.Civ.P. 54(b). Because we can consider only the issues that have been certified under Rule 54(b), Ebasco's bill of costs is not properly before us on appeal. We have no jurisdiction now to review either that bill or the trial court's ultimate award.

G. Mutual Mistake.

Finally we come to Ebasco's greatest loss under the JNOV. The trial court refused to apply the jury's finding of no mutual mistake, and held that the parties had intended the policies to contain a coordinating provision, limiting recovery for one claim under both policies to a maximum of $25 million. We have rejected the assertion that the Seventh Amendment guarantees the right to a jury on a claim of mutual mistake seeking contractual reformation.38 Such a reformation is an equitable decision made by the court, not the jury, and the parties are not entitled to a jury trial even where there is conflicting evidence of historical facts.39 Moreover, the court is free to disregard the findings of an advisory jury.40

Any substitute findings made by the court are reviewed under the clearly erroneous standard. That is, we determine only whether the court clearly erred in its finding as to whether there was clear and convincing evidence of mutual mistake.41 Here, as in Great Atlantic, "[d]espite the apparent procedural posture of this case as an appeal from an order granting a motion for [JNOV], we believe this is in fact an appeal from an order granting reformation of an insurance policy...."42

The trial court in this case specifically found "that Defendants [the insurers] proved by clear and convincing evidence that both parties intended that the policy limits would be combined in the event that a single claim arose under both policies". In support of this finding, the trial court noted the testimony of Ebasco's employee Michael Whelan (who stated that Ebasco had intended to obtain only $25 million in insurance coverage) and the discrepancy between the premiums for the two policies. Ebasco points out that Whelan did not participate in the conclusion of the negotiations for insurance, and it devalues the importance of the different premiums; it suggests that the parties considered the activities insured by the GA policy (for which the premium was almost 40 times greater) more likely to result in a claim on that policy. The insurers suggest, however, that Ebasco's coverage (for a total of $25 million under any one claim) was divided between two policies only because of reinsurance treaties (under which the underwriters could not have insured any one claim for more than $25 million); they also note other evidence, contemporaneous with the formation of the contracts,43 that confirms the intent of both parties that the policies contain a coordinating provision.

Ebasco calls attention to the $7.5 million deductible that applies to one claim arising under both policies.44 It argues that the parties would not have agreed to a higher deductible for a claim implicating bothpolicies unless the recovery would also be higher. The insurers have not refuted this point.45

After examining the record closely, we conclude that the trial court did not err clearly in finding that the insurers had presented clear and convincing evidence that they and Ebasco intended (that is, agreed) for the policies to contain such a provision, and that it was due to their mutual mistake that the policies did not contain it. There is not as much testimony in the record as we might have expected the parties (especially Ebasco, which stood to lose the most on this issue) to present. Nevertheless, no testimony by representatives of either side of the insurance policy negotiations suggest that they intended, at the time of those negotiations, to obtain, or provide, insurance coverage exceeding $25 million for any one claim. The trial court correctly reformed the policies to reflect this fact.

Ebasco argues that the trial court's finding of the parties' intent, even if correct, forms an insufficient legal basis for contract reformation based on mutual mistake. It notes that Texas law on this issue requires a finding not only of intent, but of actual agreement between the parties as to the terms they mistakenly omitted from their contract. In Champlin Oil . Chastain, the Texas Supreme Court did indeed require "proof of a definite agreement between the parties which has been misstated in the written memorandum because of a mistake common to both contracting parties".46 This Court interprets Texas law to require a similar showing, "that the parties actually agreed to a certain term".47 Yet Champlin also equates "a definite agreement" with "a meeting of the minds",48 suggesting that the requirements of intention, agreement, and mutuality are simply different aspects of the same test. And at least some Texas courts appear to consider a showing of intention and a showing of a "true agreement" as two expressions of the same requirement.49 Although we can imagine cases where the intent to agree might not actually become an agreement between the parties, we hold that the trial court's finding of intent in this case was legally sufficient to satisfy the requirement of Texas law that a court only reform a contract if there is a true agreement which the parties both intended their contract to embody. What the trial court found was that the parties mutually intended the insurance policies to contain a "tying" or coordinating clause limiting recovery for a single claim, even where it implicated both policies, to a maximum of $25 million. If they did both so intend, then their intent would represent a true, a definite, and an actual agreement. In this case it does not matter if one reaches that answer by focusing, as the trial court did, on the clear intent that gives substance to such an agreement or, as Ebasco would have had it do, on the agreement based on that intent.

The jury below gave Ebasco (and the MDL bondholders who will receive much of any award against the insurers) a great victory. The trial court took away the financial bulk of that victory. We have given some things back to Ebasco, and kept others beyond its reach.

We have found, as the trial court did, that each insurer owed Ebasco a duty to defend. Because they breached that duty the trial court properly denied them discovery into the settlement Ebasco reached with the MDL plaintiffs. The trial court also properly denied them any additional defenses based on alleged warranties or misrepresentations. Only as to Section I(c) of the GA policy do we find that Ebasco may owe a level of proof (as to coverage) that it was not asked to provide below.

We raise the level of prejudgment interest from 6% to 10%, but we apply it to a much shorter period of time. We deny such interest on attorney fees, but we award Ebasco its attorney fees for prevailing in this action. We affirm the trial court's decision to deprive Ebasco of its in-house costs in the MDL suit. We also affirm the trial court's application of the deductible endorsement to both policies. We cannot consider the trial court's treatment of Ebasco's bill of costs. We affirm the trial court's equitable reformation of the insurance contracts to a coordinated maximum recovery of $25 million for one claim, and we remand for an apportionment of covered damages.

We realize that we leave unanswered several difficult questions in what has become one prolonged case on top of another. We hope that this decision, and any further proceedings in this case, will not unduly reward insurers who wrongfully breach their duties to defend or insureds who seek indemnification for actions they never insured. Treading the fine line between these two errors will not be easy; treading it at all would have been unnecessary had the insurers in this case come forward with a defense of their insured at the appropriate time. We AFFIRM in part, REVERSE in part, and REMAND.

ON PETITION FOR REHEARING AND SUGGESTION FOR REHEARING EN BANC

March 9, 1992.

The Petition for Rehearing is DENIED and no member of this panel nor Judge in regular active service on the Court having requested that the Court be polled on rehearing en banc, (Federal Rules of Appellate Procedure and Local Rule 35) the Suggestion for Rehearing En Banc is DENIED.

To clarify the holding of the Court, the last sentence of Section III D of the February 14, 1992 opinion, is deleted and the following sentence is substituted for the deleted sentence: