United States Court of Appeals
                     For the First Circuit

No. 11-1064

                    UNITED STATES OF AMERICA,

                            Appellee,

                               v.

                    MOISÉS CANDELARIA-SILVA,

                      Defendant, Appellant.


          APPEAL FROM THE UNITED STATES DISTRICT COURT
                 FOR THE DISTRICT OF PUERTO RICO

         [Hon. José Antonio Fusté, U.S. District Judge]


                             Before

                   Torruella, Selya and Lipez,
                         Circuit Judges.


     Judith H. Mizner, Assistant Federal Public Defender, District
of Massachusetts, for appellant.
     Justin Reid Martin, Assistant United States Attorney, with
whom Rosa Emilia Rodríguez-Vélez, United States Attorney, Nelson
Pérez-Sosa, Assistant United States Attorney, Chief, Appellate
Division, and Julia M. Meconiates, Assistant United States
Attorney, were on brief for appellee.




                          May 13, 2013
           TORRUELLA, Circuit Judge.      This is a second appeal from

the   denial   of   Defendant-Appellant   Moisés   Candelaria-Silva's

("Moisés") motion for reduction of sentence under 18 U.S.C. § 3582

(c)(2) and the retroactively amended crack-cocaine guidelines. The

first time around, the district court found defendant ineligible

for a reduction, stating, without further explanation, that "[a]ny

of the other narcotics [underlying this offense] standing alone

substantiate[s] the [Offense Level] of 42 for which the defendant

was convicted."     United States v. Candelaria-Silva, 

 (1st Cir. 2009) (per curiam).       Because the court's conclusion

was not self-evident on the face of the record, we vacated that

order on appeal and remanded for further proceedings. In so doing,

we made plain that, under the applicable law, Moisés "may only be

held responsible for those drugs he personally handled as well as

those that were reasonably foreseeable to him."       Id. at 307.

           The district court reaffirmed its ruling on remand.      It

found that the quantity of heroin distributed by the conspiracy, of

which Moisés was a member, was reasonably foreseeable to him and in

itself sufficient to support the sentence imposed. Moisés contends

that the district court's foreseeability finding and the drug

quantity determinations underlying it were clearly erroneous.       We

agree.




                                  -2-
                              I. Background

           Moisés was arrested in February 1995, along with more

than 30 co-conspirators, for his part in a massive drug conspiracy,

which had controlled a substantial share of the Puerto Rican drug

market since at least 1988.           At the height of its power, the

conspiracy   headed    by    Israel    Santiago-Lugo    ("Santiago-Lugo")

controlled the drug trade in the northern half of Puerto Rico,

generating millions of dollars in profits and waves of violent

reprisals against their competitors.

           Though he was more than a street-level dealer, Moisés was

a comparatively young, minor player in the conspiracy.              Evidence

presented at trial supported a conclusion that for at least some

period of time, Moisés controlled the conspiracy's drug point at

Villa Evangelina, a public housing project in Manatí, Puerto Rico.

Exactly when and for how long Moisés controlled Villa Evangelina is

not clear.      One former co-conspirator testified that he thought

Moisés was working at Villa Evangelina in 1992 or 1993, but could

not be sure.1

           Beyond his involvement at Villa Evangelina, however, no

evidence   presented    at   trial     explicitly   tied   Moisés   to   the

conspiracy before 1992 or 1993.             Testimony of co-operating co-

conspirators at trial indicated that Moisés' mother and his older



1
  Moisés apparently became the head of the Villa Evangelina point
after the incarceration of his older brother in 1993.

                                      -3-
brother Eulalio had been stashing and packaging drugs in the

Candelaria-Silva's family home for several years, but no evidence

of record directly linked Moisés to those activities.

           As relevant here, at trial, the government presented two

ledgers that the police had seized from a co-conspirator during a

search of an apartment in Virgilio-Dávila.            An FBI cryptanalysis

expert testified that the ledgers spanned from October 1990 to

October 1991,   and    detailed,   in    units,    the     quantity   of drugs

supplied over that time-span by the Santiago-Lugo organization to

several drug points.    No mention of Moisés, his family members, or

Villa Evangelina was made in the ledgers.

           Co-operating    co-conspirator         Marcos     Hidalgo-Meléndez

("Hidalgo"), who had been in charge of cocaine distribution in Los

Murales, testified to the use of the ledgers by the Santiago-Lugo

organization as well as to the quantities of drugs reflected

therein.   Another co-operating co-conspirator, Carlos Otero-Colón

("Otero"), testified that he delivered cocaine to the Candelaria-

Silvas to be packaged before it was sold at Hidalgo's point in Los

Murales. He also testified that, "at some point," after one of his

deliveries to the Candelaria-Silvas, he attempted to open his own

drug point in Vega Baja, Puerto Rico, and that his transactions in

relation to those efforts were recorded in the ledgers.

           The jury found Moisés guilty of conspiracy to possess

with the intent to distribute fifty grams or more of cocaine base,


                                   -4-
five kilograms or more of cocaine, one or more kilograms of heroin,

and an undetermined quantity of marijuana, in violation of 21

U.S.C. §§ 841 and 846.2     Moisés was also found guilty of possession

with intent to distribute cocaine base, cocaine, heroin, and

marijuana.    At the sentencing hearing, the district court found a

base offense level of 38, and added two two-level enhancements due

to Moisés' use of a firearm and his role as a supervisor, resulting

in a total offense level of 42. He received a 30-year incarcerative

sentence.

            Following    the      2007     Amendments   to   the   Sentencing

Guidelines,   which     reduced    the     crack/powder   disparity,   Moisés

petitioned for re-hearing and was denied without explanation.              He

appealed to this court, and in a per curiam opinion, we remanded to

the district court, with instructions to provide an explanation for

its conclusions.

            On remand, the district court concluded that Moisés was

not entitled to a sentencing reduction because of evidence on the

record supporting a conclusion that he had possessed enough heroin

to warrant a base offense level of 38, regardless of any change in

the crack-related guidelines. The district court reached this

finding through combining two pieces of evidence in the record -–



2
   All co-conspirators who opted to go to trial were also found
guilty. Some, including Moisés, appealed their convictions, which
we affirmed. See United States v. Candelaria-Silva, 

(1st Cir. 1999).

                                         -5-
the ledgers recovered by the police and trial testimony explaining

the contents of the ledgers.

          According to the testimony of the FBI's cryptanalysis

expert, the ledgers recorded the sale of 28,208 units of 'c', 7,802

units of 'r', 753 units of 'a', and 9,535 unidentified units. Co-

conspirator Hidalgo testified that the 'c' was heroin and 'r' was

cocaine and that the units were packets. He further testified that

there were 50 packets in 1/8 of a kilogram of cocaine (or 400

packets in a kilogram).

          Hidalgo's critical testimony relating to the quantity of

heroin in a packet, however, was less clear and possibly marred by

prosecutorial error.   Hidalgo testified that 100 packets of heroin

were sold at a certain drug point every week.   The prosecutor then

asked him, without foundation:

          Q: Did you nevertheless find out how much he
          would pay for that eighth of a kilogram that
          you previously stated was sold every week at
          the Los Murales housing project?

          A: I was aware, I had knowledge, that at that
          point in time the eighth of a kilo of heroin
          was being sold in the market for $28,000.

          Even though Hidalgo's answer assumed a fact not otherwise

in evidence, the district court decided to credit this response as

an affirmation of the prosecutor's statement that 1/8 of a kilogram

of heroin was being sold at Los Murales every week.   Combining this

with Hidalgo's previous testimony that 100 packets had been sold

every week, the district court concluded that 100 packets of heroin

                                 -6-
equaled 1/8 of a kilogram.     The court reached this conclusion

despite noting that Hidalgo had earlier testified that each packet

sold for $75, which, assuming a $28,000 market price for 1/8 of a

kilogram of heroin, would suggest that there were closer to 400

packets in 1/8 of a kilogram. Indeed, the district court explained

in a footnote that it believed that Hidalgo had made a mistake in

his testimony and that he had meant to say that one kilogram of

heroin sold for $28,000.

          Using the 100 packets as an equivalency for 1/8 of a

kilogram of heroin, the district court divided the 28,208 units of

'c' by 100 to get the number of kilograms sold, and then divided

again by eight to conclude that for the period October 1990 -

October 1991, the conspiracy-wide total quantity of heroin was

35.26 kilograms.

          From the testimony of co-conspirator Otero that he had

delivered drugs to the Candelaria-Silva's family home at some

point, which may have been during the period reflected in the

ledgers, the district court concluded that Moisés had been involved

in the conspiracy at this time and the entire quantity of drugs was

"reasonably foreseeable" to him.

          Reasoning that 30 kilograms of heroin alone is enough to

trigger a base offense level of 38 "and because it was obvious to

all involved that the immense quantities of drugs distributed over

this seven-year conspiracy justified a base offense level of


                               -7-
thirty-eight," the court concluded that Moisés was not eligible for

a sentence reduction.   This appeal promptly ensued.

                           II. Discussion

          The two-step analysis a district court uses to determine

whether to   grant   a sentence    reduction under    §   3582(c)(2) is

straightforward. See, e.g., Dillon v. United States, ___ U.S. ___,

, 2691 (2010).     The court begins by determining "the

prisoner's eligibility for a sentence modification and the extent

of the reduction authorized."     Id.    At this first stage, the court

considers whether it has the legal authority to grant the reduction

requested; thus, its conclusions of law are reviewed de novo, and

its factual findings, for clear error.           See United States v.

Fanfan, 

, 107 (1st Cir. 2009); see also       United States

v. Davis, 

, 1303 (11th Cir. 2009); United States v.

Johnson, 

, 623 (6th Cir. 2009).        Next, at the second

step, the court determines "whether the authorized reduction is

warranted, either in whole or in part, according to the factors set

forth in [18 U.S.C.] § 3553(a)."         Dillon, 130 S. Ct. at 2691.

Decisions at this stage are reviewed for abuse of discretion, as

the question whether to reduce a final sentence pursuant to § 3582

(c)(2) "is a matter [Congress] committed to the sentencing court's

sound discretion."   United States v. Aponte-Guzmán, 

,

159-61 (1st Cir. 2012).




                                   -8-
            Moisés'      appeal   revolves around       the   factual    findings

underlying the district court's eligibility determination.                     The

clearly erroneous standard is therefore the compass that guides our

review.     The scope of our task is well settled: a reversal on

clearly erroneous grounds is in order "when . . . the reviewing

court on the entire evidence is left with the definite and firm

conviction that a mistake has been committed."                United States v.

United Gypsum Co., 

, 395 (1948).            Such is the case, for

example, when the district court fails "to synthesize the evidence

in a manner that accounts for . . . gaps in a party's evidentiary

presentation."      Doe v. Menefee, 

, 164 (2d Cir. 2004).

Likewise, reversal for clear error is warranted "where the trial

court incorrectly assessed the probative value of various pieces of

evidence, leading it to rely on speculation . . . ."                 Id. (citing

United States v. Rizzo, 

, 100-02 (2d Cir. 2003)); see

also United States v. Marrero-Ortiz, 

, 779-80 (1st Cir.

1998)     ("While   we    may     agree    with   the   government      that   the

[conspiracy] did a substantial amount of narcotics business, and

that the totals necessary for a [drug quantity finding] seem

attainable given the appellant's role in the conspiracy, we cannot

uphold a drug quantity calculation on the basis of hunch or

intuition.").

            Applying this standard to the record before us, we find

clear error in the district court's factual finding that the


                                          -9-
quantity   of   heroin   reflected   in   the   ledgers   was   reasonably

foreseeable to Moisés.    The district court reached that conclusion

without addressing troublesome evidentiary gaps, all of which

Moisés underscores.

           For example, among other things, the ledgers nowhere

mention (either explicitly or in code) Moisés, his family, or the

drug points attributed to them.       In fact, trial testimony from a

government's expert established that not a single ledger entry

could be attributed to Moisés.        Similarly, no evidence of record

shows that Moisés participated in any way in the preparation of the

ledgers.   Nor is there any evidence directly linking Moisés, his

family, or the Villa Evangelina project to the co-conspirators from

whom the ledgers were seized.        And the housing project where the

seizure occurred was a long distance away from the Candelaria-

Silvas' home base in Manatí.         Last but not least, the ledgers

covered transactions occurring between 1990 and 1991, a period

during which Moisés' brother was the leader of the Candelaria-

Silvas' drug operation.       Although Moisés apparently took over

business at Villa Evangelina after his brother's arrest in 1993, no

evidence of record shows what Moisés' role in the conspiracy was

when the transactions recorded in the ledgers took place.          We thus

agree with Moisés that "there is nothing in the contents of the

ledgers or the circumstances of their seizure to suggest that [he]




                                 -10-
was aware of them or to support attributing the quantities they

referenced to him."

          The government interprets the record differently.             From

its vantage point, "the record establishes that Moisés was a high

member of the Santiago-Lugo organization [who] fully participated

in the drug trade . . . [and enjoyed] a position of trust and

cooperation with other members of the . . . drug conspiracy."               The

government offers four factual assertions in support: (1) that

Moisés became the leader of the Villa Evangelina drug point after

his brother's arrest in 1993; (2) that the Candelaria-Silvas

packaged drugs received from Virgilio-Dávila for sale in Los

Murales; (3) that Santiago-Lugo himself offered protection to

Moisés in connection with a personal feud with people from another

residential project;3 and (4) that Santiago-Lugo's brother, who was

married to Moisés' sister, would help from time to time at the

Villa Evangelina drug point.        The government's factual proffer,

however, falls far short of satisfying the applicable burden.

          For sentencing purposes in a drug-distribution conspiracy

conviction,   aside   from   his   or   her   own   acts,   a   defendant    is

accountable only for "all reasonably foreseeable quantities of

contraband that were within the scope of the criminal activity that



3
   At some point during the drug conspiracy, Moisés was kidnaped
and robbed while visiting a drug point at a neighboring residential
project. Santiago-Lugo thereafter told Moisés that he would be
nearby if needed.

                                   -11-
he jointly undertook."       U.S.S.G. § 1B1.3, cmt. (n.2).           This means

that the sentencing court must "ascertain on an individual basis

the scope of the criminal activity that the particular defendant

agreed jointly to undertake."            United States v. Carrozza, 

, 76 (1st Cir. 1993); see also United States v. Cruz-Rodríguez,

, 32 (1st Cir. 2008) ("When making the individualized

finding    of   drug    quantity   responsibility,     the   court    must   not

automatically shift the quantity attributable to the conspiracy as

a whole to the defendant.").         Specifically, the record must show

the defendant's "level of involvement so as to explain why the

nature of the conspiracy or his relationship with the leaders of

the conspiracy showed he could foresee a given quantity of drugs."

United    States   v.    Correy,   

,   388   (1st   Cir.   2009)

(emphasis in original).

            The immense size of Santiago-Lugo's organization was a

matter thoroughly discussed during the trial against Moisés and its

other members.         At Moisés' sentencing hearing, the size of the

conspiracy played a major role in the court's analysis: "[a]fter

considering the immense size of this drug-trafficking conspiracy,

comprising so many members and various kinds of drugs distributed

throughout the northern half of the island, it was not difficult

for the experienced eye of the trial judge to arrive at a [base

offense] level of thirty-eight."            The record shows that, within

Santiago-Lugo's vast drug-trafficking network, the Candelaria-


                                     -12-
Silvas played a discrete role through their activities at Villa

Evangelina and Los Murales.       The record also shows that in 1993

Moisés may have succeeded his brother as the leader of the Villa

Evangelina drug point.      But other than these facts, the record is

devoid of any evidence from which we could explain why Moisés'

seemingly     discrete     role   within   Santiago-Lugo's      "immense"

organization put him in a position to foresee the quantities of

drugs handled by it.       See United States v. Willis, 

,

1274 (7th Cir. 1995) ("[I]t is highly questionable to leap from one

person's knowledge that the organization is big to knowledge of its

full scope.     The district judge must take a closer look at this

subject.").    Nor does the record show that Moisés was privy to any

information from which he could foresee the drug quantities that

Santiago-Lugo's organization handled.      While Moisés and Santiago-

Lugo appear to have had some personal ties, more is necessary to

show that the drug quantities involved in the conspiracy were

foreseeable to Moisés.      See Correy, 570 F.3d at 388 (noting that a

drug-conspiracy    sentence    premised solely   upon   the   defendant's

familiarity with the leaders of the conspiracy "goes against our

instruction to base individualized drug finding on a review of the

record").     The government's proffer does little to address these

types of concerns.       We thus discard its contentions on this front

as insufficient.




                                   -13-
           Similar problems afflict the district court's factual

conclusions about the quantity of drugs recorded in the ledgers.

We have stated on previous occasions that where, as here, a drug

quantity   determination     relies   on    multiples   of    averages    or

extrapolations, the sentencing court must be mindful of "the

potential for error where one conclusory estimate serves as the

multiplier for another (i.e., average number of transactions per

hour and average operating hours per day) [, which] may undermine

the reasonable reliability essential to a fair sentencing system."

United States v. Rivera-Maldonado, 

, 233 (1st Cir.

1999); see also United States v. Sepúlveda, 

, 1198 (1st

Cir. 1993) ("[T]he two flawed findings feed on each other; by using

not one, but two, unsupported averages to arrive at both the number

of trips undertaken and the amounts of cocaine handled in the

course of each trip, the court compounded the error of its ways.").

Accordingly, in sentencing a defendant convicted of participation

in a poly-drug conspiracy, "care must be taken to ensure that

particularized   drug-type    quantity     findings   are    predicated   on

reliable information and, where significant uncertainty exists,

that those findings err on the side of caution." Rivera-Maldonado,

194 F.3d at 233-34 (concluding that drug quantity determination was

clearly erroneous where "the risk of error was compounded by

pyramiding unreliable inferences").         In other words, sentencing

judges may rely "on reasonable estimates and averages, "id. at 228,


                                  -14-
but not    on   drug   quantity   calculations   based   on   "hunches   and

intuition,"     Marrero-Ortiz, 160 F.3d at 779-80.

           Here, the district court based its determination that

there were 100 "packets" in 1/8 of a kilogram of heroin on

testimony that assumed a fact not in evidence, that the district

court acknowledged was inconsistent with prior testimony, and that

the district court concluded was probably actually mistaken. These

numbers are not the sort of "reasonable estimates and averages"

that can or should be used as the foundational multiplier when

making a drug-quantity determination.       The risk of error inherent

in these loose calculations is simply too high.               As such, we

conclude that the district court's drug quantity calculation was

also clearly erroneous.

           The government argues in the alternative that, even if

Moisés was eligible for a sentence reduction, the district court

would have found the § 3553(a) factors to preclude the relief

sought.    The district court, however, explicitly declined to

consider   § 3553(a) given its ineligibility finding; therefore, we

are not in a position to make any determinations in this regard.

See Aponte-Guzmán, 696 F.3d at 159-61 (stating that the balancing

of § 3553(a)'s factors is "committed to the sentencing court's

sound discretion"); see also United States v. Cardosa, 

,

22 (1st Cir. 2010) ("Cardosa is eligible for resentencing; whether




                                    -15-
to do so is within the discretion of the district judge on

remand.").

                               III. Conclusion

           For the reasons stated above, we vacate the court's

judgment and remand for further proceedings.             In so doing, we note

that despite the considerable amount of resources and time spent in

addressing   Moisés'    motion,      the    record    twice   presented   to    us

contains scant evidence from which to conclude that Moisés is

ineligible for the relief requested.               On remand, therefore, the

parties as well as the court would be well advised to move beyond

the eligibility question and squarely address the second step of

the   applicable    analysis    --   that    is,     determining   whether     the

"reduction is warranted, either in whole or in part, according to

the factors set forth in [18 U.S.C.] § 3553(a)."                Dillon, 130 S.

Ct. at 2691.

           We have the utmost confidence in the district court

judge's   ability      to   adjudicate       Moisés'     motion    fairly      and

objectively.       We understand that district court judges retain

considerable discretion in fashioning an explanation of their

sentencing decisions.       However, given the unusual circumstances of

this case (a second remand because of errors in the district

court's handling of the resentencing decision), we cannot emphasize

more strongly the importance of the district court's duty to




                                      -16-
provide   detailed   support   for   both   its   factual   and   legal

conclusions.

          So Ordered.




                                -17-