In this case, the district court departed downward from the guideline sentencing range (GSR) as to each of five defendants on the theory that the harm attributed to them, measured by the amount of loss sustained by the victim, overstated the seriousness of the offense of conviction. The government now asks us to evaluate both the lawfulness of the downward departures and the propriety of the court’s role-in-the-offense adjustments for two defendants, David and Steven Ros-toff. We uphold the sentences of all defendants except the Rostoffs (who must be re-sentenced as a result of erroneous role determinations).
I. BACKGROUND
A federal grand jury indicted the brothers Rostoff, together with James Harris, Dolores DiCologero, and Paul J. Bonaiuto, on charges, inter alia, of conspiracy, bank fraud, and the making of false statements. See 18 U.S.C. §§ 371,1344, and 1044. These *403charges stemmed from a failed foray into the New England condominium market — a market that rose to giddy heights in the mid-to-late-1980s and then plunged precipitously.
The conspiracy count constituted the hub of the indictment. In it, the grand jury charged that, from December 1985 to February 1989, the defendants, aided and abetted by others, fraudulently induced a federally insured financial institution, the Bank for Savings (the bank), to grant several hundred loans, totalling in excess of $30,000,000, to persons purchasing condominium units from David Rostoff, Steven Rostoff, and James Harris (collectively, “the Rostoff group” or “the developers”). Like spokes running from the hub, 43 of these loans gave rise to 86 “mirror image” bank fraud and false statement counts against various defendants.
The trial jury plausibly could have found that the scheme tracked the following script. The bank had a firm policy of refusing to grant first mortgage loans in excess of 80% of the lower of the sale price or the appraised value of residential real estate; and, when mortgages were written on that basis, the bank ordinarily required the balance of the purchase price to be paid in cash by the borrower. In 1986, bank officials, eager to maintain a lucrative working relationship with the Rostoff group, bent the rules. The bankers allowed the developers to assist common customers (i.e., persons who bought condominiums from the Rostoff group and financed the purchases through the bank) in an uncommon way: by taking back second mortgages to circumvent the cash down-payment requirement. The bankers conditioned this concession on the express understanding that the second mortgages would be enforced, and that each purchaser would make at least a 10% down payment from his or her own capital.
This arrangement proved too tame for the developers’ purposes. To facilitate sales, they cooked the books, surreptitiously telling selected buyers that they would not enforce the second mortgages, or, alternatively, that they would not demand interest payments on particular second mortgages until resale of the encumbered condominiums. More importantly, the developers set out to subvert the down-payment requirement by orchestrating a paper shuffle designed to create the (false) impression that the buyers were putting 10% down in order to acquire the properties, when they were not. In many instances, the developers accomplished this sleight of hand by offering customers a 10% discount from the stated purchase price. When a customer agreed to buy at the reduced price, the developers submitted documents to the bank that overstated the actual purchase price by 10% and treated the negotiated discount as a down payment. This flim-flam took on added significance because the bank underwrote the loans on the basis of an 80% loan-to-value (LTV) ratio, using purchase price as a principal measure of value. Thus, an inflated purchase price often caused the bank to approve a higher, first mortgage loan than would have been forthcoming had it known the true purchase price. In the end, many buyers acquired condominiums without making any down payment or other cash expenditure (except for closing costs).
The bank’s closing attorney, defendant Bo-naiuto, and the manager of the bank’s mortgage department, defendant DiCologero, knowingly participated in fabricating this tissue of lies, half-truths, and evasions. Between September 1986 and February 1989, the bank engaged Bonaiuto to close at least 240 loans to the developers’ customers. Although no fewer than five borrowers testified at trial that they asked Bonaiuto about apocryphal deposits shown on their settlement sheets, he did not notify the bank of any discrepancies.1 DiCologero also worked closely with the developers, handling the day-to-day administration of the loan approval *404process. The prosecution proved her awareness of the ongoing scheme largely by circumstantial evidence.2
Following a lengthy trial, a jury found each of the five defendants guilty of conspiracy to defraud the bank. In addition, the jury found Steven Rostoff guilty on a total of 72 “mirror image” counts of bank fraud and making false statements (representing 36 transactions), David Rostoff guilty on 32 such counts (representing 16 transactions), Harris guilty on 52 such counts (representing 26 transactions), Bonaiuto guilty on 10 such counts (representing five transactions), and DiCologero guilty on two such counts (representing one transaction).
On January 29, 1993, the district court convened a disposition hearing.3 By then, the bank had become insolvent, and the Federal Deposit Insurance Corporation (FDIC) had become the receiver. The court determined that the FDIC sustained losses due to the defendants’ activities in the $2,000,000-$5,000,000 range. The court then calculated the offense level of all defendants except DiCologero on the basis of this loss computation, see U.S.S.G. § 2F1.1(b)(1)(E) (providing a 10-level enhancement for fraud crimes involving losses of more than $2,000,000, up to and including $5,000,000), arriving at an adjusted offense level (OL) of 20 for the Rostoff brothers and Bonaiuto, and 18 for Harris. The court attributed slightly under $1,000,-000 in losses to DiCologero and, after other interim adjustments, settled on an OL of 18. The court factored in the defendants’ criminal history scores — all were first offenders— and arrived'at a GSR of 33-41 months at OL-20 and a GSR of 27-33 months at OL-18. Finding, however, that in each instance the amount of loss overstated the seriousness of the particular defendant’s criminality, Judge Zobel departed downward. She sentenced David and Steven Rostoff to serve 15-month terms of immurement; sentenced Harris to a nine-month prison term; sentenced Bonaiuto to two years probation, six months of which was to be served in a community treatment center; and sentenced DiCologero to two years of straight probation. This appeal followed.
II. THE DOWNWARD DEPARTURES
In sentencing under the guidelines, departures are the exception rather than the rule. See United States v. Diaz-Villafane, 874 F.2d 43, 52 (1st Cir.), cert. denied, 493 U.S. 862, 110 S.Ct. 177, 107 L.Ed.2d 133 (1989). When a district court nonetheless departs, and an appeal eventuates, we ask three general questions: (1) Is the reason that the sentencing court gave for departing of a type that lawfully can ground a departure in an appropriate case? (2) Is the court’s factfinding in respect to the cited reason sustainable on whole-record review? (3) Is the degree of the departure reasonable? See United States v. Mendez-Colon, 15 F.3d 188, 189 (1st Cir.1994); United States v. Rivera, 994 F.2d 942, 950-52 (1st Cir.1993); Diaz-Villafane, 874 F.2d at 49. A departure passes muster only if all three inquiries yield an affirmative response.
In this case, the government asserts that the lower court erred at each step along the *405departure path. We trace the contours of the court’s decision and then address the three relevant questions.4
A. The Anatomy of the Departure Decision.
In fraud cases controlled by the guidelines, the amount of the victims’ monetary loss (actual or intended) is a proxy for the seriousness of the offense, and, thus, a key determinant of the severity of the perpetrator’s sentence. See United States v. Lilly, 13 F.3d 15, 17, 19 (1st Cir.1994); United States v. Tardiff, 969 F.2d 1283, 1285 (1st Cir.1992). Recognizing, however, that no proxy is perfect, the applicable edition of the sentencing guidelines cautions that:
In a few instances, the total dollar loss that results from the offense may overstate its seriousness. Such situations typically occur when a misrepresentation is of limited materiality or is not the sole cause of the loss ... In such instances, a downward departure may be warranted.
U.S.S.G. § 2F1.1, comment, (n. 11) (Nov. 1987). ■
The defendants in this case all moved for downward departures based on application note 11. The district court accommodated their requests, linking its largesse to a linchpin finding that numerous factors, apart from the defendants’ conduct, inflated the losses sustained by the FDIC. The court premised its linchpin finding primarily on three subsidiary findings. (1) The court remarked the bank’s gadarene rush to participate in the condominium boom despite the obvious risks. To the court’s way of thinking, this overeag-erness was driven by greed — after all, the bank based incentive compensation for top officials on loan production and fomented a “lend at all costs” mentality that led senior managers to condone the defendants’ shenanigans. The court expressed great skepticism about senior management’s professed lack of knowledge or responsibility, concluding that, at the very least, management had acted negligently, particularly in authorizing loan approvals, and had bent its policies grotesquely to retain the Rostoff group’s business. In the court’s view, these shortcomings contributed mightily to the extent of the eventual losses. (2) Next, the court found that the buyers were neither dupes nor victims in the traditional sense. To the contrary, the court thought they had become willing participants in the defendants’ scheme. Their cupidity drove up prices in the condominium market and, thus, contributed substantially to the amount of money eventually lost. (3) Finally, the court observed that economic forces not under the control of, or precipitated by, the defendants, especially the sudden, unforeseen collapse of the New England real estate market — a collapse that decimated the demand for residential condominiums — increased the magnitude of the losses.
The district court believed that these factors, in combination, contributed so directly to the extent of the loss that the defendants were entitled to a substantial measure of relief. In the sections that follow, we test the legal and factual sufficiency of the court’s stated ground. Finally, we examine the reasonableness of the actual departures insofar as they affect Harris, DiCologero, and Bonai-uto.
B. Step One: The Court’s Reason.
While the government assails the departure decision on all available fronts, its fundamental point is that, as a matter of law, the guidelines simply do not authorize departures under a “multiple loss causation” theory. Since this assertion questions whether the departure-justifying reason cited by the court below is of a kind that the guidelines, in principle, permit a sentencing court to embrace for that purpose, we afford plenary review. See Rivera, 994 F.2d at 951; Diaz-Villafane, 874 F.2d at 49.
In evaluating multiple loss causation as a departure-justifying circumstance, we do not write on a pristine page. In United States v. Gregorio, 956 F.2d 341 (1st Cir.1992), we approved the manner in which the district court, acting under the general fraud guideline, U.S.S.G. § 2F1.1, structured its down*406ward departure to “reflect[ ] ‘multiple causation’ for victim loss.” Id. at 344. Although the “sufficiency of the basis for departing in response to multiple causation of victim loss” was not at issue on that occasion, id. at 347 n. 10, we stated unambiguously that “ ‘multiple causation’ of victim loss is a ‘Commission-identified’ circumstance in which a downward departure may be warranted.” Id. at 347. We do not believe that these words, even if technically dictum, can be read other than as an outright endorsement of multiple loss causation as a permissible basis for departing downward, and, indeed, as a departure-justifying reason that the guidelines encourage. See generally Rivera, 994 F.2d at 948 (explaining that the guidelines sometimes “offer the district court, which is considering whether to depart, special assistance, by specifically encouraging” certain types of departures).
Despite the plain import of Gregorio, the government maintains that multiple loss causation is an invalid basis for a downward departure. Gregorio is irrelevant here, the government says, because the Gregorio court had before it the November 1990 version of the guidelines, which, like the original (1987) version, authorized departures when “the total dollar loss that results from the offense [overstates] its seriousness,” such as when “a misrepresentation ... is not the sole cause of the loss.” 956 F.2d at 345 (citing November 1990 version of application note ll).5 In the government’s view, time has passed Gregorio by, for the Sentencing Commission rewrote the application notes to section 2F1.1 effective November 1, 1991, consolidating several preexisting notes into a new note 10. In the process, the Commission eliminated any reference to “the sole cause of the loss” language.6 The government proceeds to weave a tapestry from several gossamer strands of speculation and surmise, hypothesizing that the Commission, recognizing that it had improvidently promulgated former note 11, acknowledged the error of its ways and junked the original reference. Using this hypothesis as a springboard, the government then jumps to the conclusion that the Commission, in essaying the revision, tacitly rejected multiple loss causation as an appropriate factor in the departure calculus.
We need not resolve the issue of whether the Commission, in revising the application notes in a way that dropped the “sole cause of the loss” language, intended to drum multiple loss causation out of the ranks of encouraged departures. To avoid ex post facto difficulties, courts should “normally apply [guideline] amendments retroactively only if they clarify a guideline, but not if they substantively change a guideline.” United States v. Prezioso, 989 F.2d 52, 53 (1st Cir. 1993); accord Isabel v. United States, 980 F.2d 60, 62 (1st Cir.1992). This rule stymies the government in this instance. If, on the one hand, as the government argues, the Commission’s rewriting of the application notes bars downward departures premised on multiple loss causation, then that revision cannot be applied retroactively for doing so would change the substance of the fraud guideline, U.S.S.G. § 2F1.1, as that guideline was explicated in Gregorio. See Prezioso, 989 F.2d at 54 (explaining that a new interpretation of a guideline that contradicts existing circuit precedent “alters the guideline” and, hence, constitutes a substantive change that can only apply prospectively). If, on the other hand, the revision does not bar downward departures for multiple loss causation, then the district court’s selection of multiple loss causation as its departure-justifying ground is, under Gregorio, unimpugnable.
Consequently, we hold that, under the original pre-1991 version of the guidelines— the version that controls here — the district court permissibly singled out multiple loss *407causation as a departure-justifying circumstance.7
C. Step Two: The Factual Predicate.
Since the lower court isolated a conceptually proper departure-justifying circumstance, the second step of the review process looms. At this stage, we must determine whether, on the whole record, the court sup-portably could have found that the departure-justifying circumstance actually existed. See Diaz-Villafane, 874 F.2d at 49. Because this determination implicates the court’s fact-finding, our standard of review is deferential. See id. (explaining that the findings of fact underlying a departure decision “may be set aside only for clear error”).
Aside from the defendants’ actions, the district court identified three factors that contributed to the magnitude of the loss in this ease: (1) the conduct of the bank’s senior management; (2) the buyers’ esurience; and (3) the nosedive in condominium prices. The government does not seriously dispute either the incidence of these factors or their aggravating effect upon the amount of loss.8 Instead, the government asserts that the court clearly erred in finding an overstatement because the loss figures that the court used for sentencing purposes represented only a fraction of the actual losses caused by the defendants’ criminal activity.
This argument will not wash. Calculating the amount of loss for purposes of the sentencing guidelines is more an art than a science. Courts can, and frequently do, deal with rough estimates. See United States v. Skrodzki, 9 F.3d 198, 203 (1st Cir. 1993); see also U.S.S.G. § 2F1.1, comment., n. 8 (stating that “the loss need not be precise,” so long as the court “make[s] a reasonable estimate of the range of loss, given the available information”). Hence, a party dissatisfied with the sentencing court’s quantification of the amount of loss in a particular ease must go a long way to demonstrate that the finding is clearly erroneous. See Skrodzki, 9 F.3d at 203; Tardiff, 969 F.2d at 1288.
Here, the court computed the amount of loss based on 43 loans that were specifically enumerated in various substantive counts of the indictment, plus an additional 97 loans that the Federal Bureau of Investigation (FBI) had classified as fraudulent. The court then excluded from its loss calculation for each defendant any loan that formed the basis for a specific count on which he or she had been acquitted. In restricting her computations to these 140 loans, the judge relied on an affidavit subscribed to by an FBI case agent, who reviewed the bank’s records and culled out loans for which he found “specific evidence of fraud.”
Bearing in mind the wide berth that sentencing judges must be given in determining what information is, or is not, sufficiently reliable to be used in imposing sentence, see Tardiff, 969 F.2d at 1287, we cannot say that Judge Zobel’s refusal to venture beyond these 140 loans constituted clear error — especially since the record contains only sketchy information about the origin and extent of losses on other loans. Nor can we say that the judge erred in excluding “acquit*408ted” loans. Although relevant conduct must be determined by the court, not the jury, see, e.g., United States v. Tavano, 12 F.3d 301, 306 (1st Cir.1993); United States v. Mocciola, 891 F.2d 13, 17 (1st Cir.1989), we believe the evidence here falls well short of compelling a finding that any “acquitted” loans must be included in calculating the amount of loss.
Because the record adequately supports the district court’s findings as to both multiple loss causation and amount of loss— indeed, the government has shown us nothing that casts serious doubt on the plausibility of the court’s findings in either respect— we conclude that the departure decision passes muster at step two.
D. Step Three: Reasonableness.
We come now to the final step in the review process, focusing on whether the “direction and degree of departure” are reasonable. Diaz-Villafane, 874 F.2d at 49. The government says that the district court stumbled at this step by failing to explain how it arrived at such sizable sentence reductions and by exhibiting unreasonable leniency. We turn first to the absence of a particularized explanation of how the district court determined the extent to which it would depart.
1. The Need for Findings. In United States v. Emery, 991 F.2d 907, 913 (1st Cir. 1993), we held that it is not necessary for a district court to “dissect its departure decision, explaining in mathematical or pseudo-mathematical terms each microscopic choice made in arriving at the precise sentence.” We opted instead for a pragmatic approach, recognizing the helpfulness of explanations but cautioning that “when the court has provided a reasoned justification for its decision to depart, and that statement constitutes an adequate summary from which an appellate tribunal can gauge the reasonableness of the departure’s extent, it has no obligation to go further and attempt to quantify the impact of each incremental factor on the departure sentence.” Id. This approach reflects our view that judicial discretion, sensibly exercised, is in most cases the critical determinant of the degree of departure. See United States v. Aymelek, 926 F.2d 64, 70 (1st Cir. 1991) (holding that, in respect to unguided departures, “a sentencing court need not resort at all to analogies”); Diaz-Villafane, 874 F.2d at 51-52 (disavowing any intention to reduce departure decisions to exercises in “mechanistic bean-counting”).
This approach is not discredited by cases such as United States v. Rosales, 19 F.3d 763 (1st Cir.1994). There, the district court gave only a terse summary of its reasons for departing, and offered no insight into how it settled upon the degree of departure. On appeal, this paucity of information compromised our ability to assess the departure’s reasonableness and necessitated a new sentencing proceeding. See id. at 770.
To be on the safe side, a sentencing judge should always endeavor to explain the extent of a departure. Yet judges are human, and, like other human beings, they will sometimes fail to dot every “i” and cross every “t.” When such a slip occurs and a sentencing court neglects to explain how it fixed the extent of a departure, no bright-line rule obtains. Such situations must be handled on a case by case basis. The bottom line is that we eschew a purely mechanical test — one that merely asks whether or not the sentencing court has made findings explaining the degree of departure — in favor of a practical one — one that asks more broadly whether or not the sentencing court has supplied the appellate panel with sufficient information to enable it to determine the reasonableness of the departure. See, e.g., United States v. Quinones, 26 F.3d 213, 219 (1st Cir.1994) (stating that the.court of appeals will overlook the omission of an explicit explanation anent the scope of a departure “if the reasons for the judge’s choice are obvious or if an explanation can fairly be implied from the record as a whole”).
Here, unlike in Rosales, appellate review is facilitated by the sentencing court’s detailed explanation of the circumstances warranting departure. This thorough exposition is adequate to explain the departures’ extent. In particular, Judge Zobel’s founded determination that the amount of loss grossly overstated the seriousness of the defendants’ *409criminal activity weighs heavily. Precisely because the guidelines use amount of loss as a proxy for culpability in fraud cases, a supportable finding that the loss exaggerates the reality of events often is tantamount to a finding that the conventional sentencing range exaggerates a defendant’s blameworthiness, and, thus, tends to invite a corresponding downward departure. So it is here. Accordingly, while we would have preferred a more deliberate discussion of the degree of departure, “we see no purpose served in this case ... in remanding to make explicit what was implicit.” United States v. Sclamo, 997 F.2d 970, 974 (1st Cir.1993).
2. The Departures’ Extent. The second shot in the government’s sling comes closer to the mark. The three departures currently under review are substantial; as we show in the margin, the least generous of them (applicable to Harris) reduces the sentence to one-third the bottom of the GSR, and the other two departures (applicable to Bonaiuto and DiCologero, respectively) manifest even greater clemency.9 Nonetheless, we reject both the prosecution’s implicit premise that unguided departures of this magnitude are presumptively unsound and its explicit premise that the particular departures sub judice are simply unreasonable.10
We begin at bedrock. In respect to unguided departures, once the sentencing court identifies a departure-justifying circumstance and decides to act upon it, there is no algebraic formula that it can invoke to quantify the departure’s extent. Hence, determining the size of such a departure is “quintessentially a judgment call,” Diaz-Villafane, 874 F.2d at 49, of a type that the law leaves almost entirely to the sentencing court’s standardless discretion. This means, of course, that there is no single, correct, “one-size-fits-all” unguided departure; rather, in any given situation, a range of widely disparate options doubtless fall within the universe of acceptable sentencing outcomes.
Similarly, once the trial judge departs, there is no litmus test that an appellate court can employ to verify that the extent of an unguided departure is — or is not— reasonable. This stark reality, coupled with the district court’s superior knowledge of the facts and its matchless ability to detect the subtle nuances that at times distinguish cases from the mine-run, argues convincingly for a deferential approach. See Rivera, 994 F.2d at 950 (discussing desirability of deference in light of “sentencing court’s superior ‘feel’ for the case”) (quoting Diaz-Villafane, 874 F.2d at 50); see generally Bruce M. Selya & Matthew R. Kipp, An Examination of Emerging Departure Jurisprudence Under the Federal Sentencing Guidelines, 67 Notre Dame L.Rev. 1, 39-40 (1991) (explaining that, in reviewing the extent of an unguided departure, “the sentencing judge’s decision is accorded generous latitude in recognition of the court’s firsthand knowledge of the case”). We have consistently held, therefore, that appellate judges must exercise considerable restraint before disturbing the presider’s reasoned definition of the degree of depar*410ture. See Rivera, 994 F.2d at 950; Diaz-Villafane, 874 F.2d at 49-50.
To be sure, this emphasis on deference places a considerable burden on the sentencing judge. To ease the weight of this burden, the judge is entitled to expect counsel’s help. The lawyers are (or, at least, they should be) a fecund source of assistance, for they have every incentive to give the trial court the benefit of their thinking on issues in the case. Indeed, the prosecution, which has an institutional interest in seeing'that justice is done, possesses an incentive that borders on an obligation.'
Departures fit neatly within this conceptual framework. Judges must forewarn the parties of imminent departures, see Burns v. United States, 501 U.S. 129,135-39, 111 S.Ct. 2182, 2185-88, 115 L.Ed.2d 123 (1991), and, once forewarned, the prosecution and the defense become full partners with the court in the departure pavane. Given the opportunity, the parties — out of self-interest, if for no more ennobling reason— should try to aid the court in determining what degree of departure best responds to the idiosyncratic features of the specific case. A prosecutor who forfeits this opportunity is in a peculiarly poor position to protest profusely when the judge, left to her own devices, thereafter exercises her discretion as she deems best.
This brings us to a special circumstance that undermines the argument the United States advances here. Judge Zobel invited the government to make recommendations at the disposition hearing concerning the appropriate degree of departure for each defendant. The prosecutor declined the invitation, clinging stubbornly to his position that the court should not depart at all. At oral argument in this venue, the government sought to justify this maneuver by suggesting that its underlying position — its claim that the district court could not lawfully depart — somehow relieved it of any responsibility to assist the court in fixing the degree of departure. We are unpersuaded.
The court below was faced with two distinct decisions: whether to depart, and if so, to what extent. Once the court resolved the threshold issue and solicited the parties’ views on the second issue, the prosecution, given its distinctive role, could not sidestep the separate inquiry as to ’ the degree of departure merely because it disagreed with the court’s initial ruling. Counsel who lose a point can neither pout nor play the ostrich, but must move on and confront the next set of issues. See, e.g., United States v. Smolar, 557 F.2d 13, 17 (1st Cir.), cert. denied, 434 U.S. 866, 98 S.Ct. 203, 54 L.Ed.2d 143 (1977). Just as a lawyer who moves unsuccessfully for judgment as a matter of law must then give the court his suggested jury instructions on the issue or risk’ a less-than-favorable charge, so, too, a prosecutor who argues against a departure, loses, and then refuses to offer suggestions referable to the degree of departure runs a comparable risk.
In this instance, the chickens came home to roost: the district court, unable to pry a recommendation out of the prosecution, granted sizable sentence reductions. Under these straitened circumstances, the government has an especially hard row to hoe in its effort to convince us that the district court displayed unreasonable generosity in shaping the departures. Because reasonableness is not an absolute, but a construct that “depends on the circumstances,” Cotto v. United States, 993 F.2d 274, 280 (1st Cir.1993), the government’s silence in the face of the lower court’s timeous request for enlightenment concerning the appropriate extent of the departures affects our assessment of the departures’ reasonableness. Put another way, the government, having been afforded an opportunity to influence a discretionary decision and having chosen instead to stonewall, can expect that doubts will be resolved against it when, thereafter, it attempts to second-guess that decision.11 Cf. Paterson-Leitch Co. v. Mas*411sachusetts Mun. Wholesale Elec. Co., 840 F.2d 985, 989 (1st Cir.1988) (“Courts, like the Deity, are most frequently moved to help those who help themselves.”).
Against this backdrop, we conclude that the government has not shown the sentencing outcomes in this case to be beyond the realm of reason. In reviewing wpward departures, we have ratified very dramatic deviations from tabulated sentencing ranges so long as they have been shown to be responsive to the record. In Diaz-Villafane, for instance, we affirmed a 120-month sentence though the GSR topped out at 33 months. In approving this upward departure — representing a 264% increase in the defendant’s sentence — we deferred to “the district court’s firsthand knowledge of the case and its careful exposition of the reasons why it thought the situation to be markedly atypical.” 874 F.2d at 52. Diaz-Villafane is not an aberration. See, e.g., United States v. Hernandez Coplin, 24 F.3d 312, 316 (1st Cir.) (upholding as reasonable 38-month and 46-month upward departures, representing increases of 380% and 328% over the respective GSR ceilings), cert. denied, — U.S. -, 115 S.Ct. 378, 130 L.Ed.2d 328 (1994); United States v. Doe, 18 F.3d 41, 48-49 (1st Cir. 1994) (upholding as reasonable a 45-month upward departure that represented a 166% increase over the GSR’s apex); United States v. Figaro, 935 F.2d 4, 8-9 (1st Cir. 1991) (upholding as reasonable an upward departure that tripled the defendant’s sentence); United States v. Rodriguez-Cardona, 924 F.2d 1148, 1156-57 (1st Cir.) (upholding as reasonable an 84-month upward departure that represented an increase of 165% over the GSR’s apex), cert. denied, 502 U.S. 809, 112 S.Ct. 54, 116 L.Ed.2d 31 (1991).
Because we do not visualize departures as a one-way street leading invariably to higher sentences, the same reasoning applies ex proprio vigore to downward departures. This street runs both ways. Consequently, the amount of deference that is due to a district court’s decision regarding the degree of departure does not expand and contract depending upon the departure’s direction. See, e.g., United States v. White Buffalo, 10 F.3d 575, 577-78 (8th Cir.1993) (upholding as reasonable a downward departure to a term of probation as against a GSR of 18-24 months); United States v. One Star, 9 F.3d 60, 61-62 (8th Cir.1993) (upholding as reasonable a downward departure to a term of probation as against a GSR of 33-41 months); Sclamo, 997 F.2d at 972 (upholding as reasonable a downward departure to a term of probation as against a GSR of 24-30 months); United States v. Jagmohan, 909 F.2d 61, 65 (2d Cir.1990) (affirming district court’s downward departure from GSR of 15-21 months to a term of probation).
We will not primp the peacock’s plumage. Here, four critical factors militate against a holding that the departures are unreasonably steep: (1) the district court’s supportable finding that the amount of loss vastly overstated the defendants’ culpability, (2) the combined impact of the several external elements cited by the court {e.g., the greed displayed by the lender’s senior management, the bank’s negligence, the buyers’ complicity, and the market’s collapse), (3) the special circumstance that the government refused to assist the court in the daunting task of determining the departures’ size, and (4) the breadth of the court’s discretion in this area of sentencing. Though the question is close, we conclude that the three departures are all within, albeit tiptoeing along the outer periphery of, the universe of acceptable sentencing outcomes.
Finally,, we think that the differences in the degrees of departure as among the various defendants are sustainable. As *412we have repeatedly observed, the amount of loss is a proxy for the seriousness of an offense. In a broad sense, then, the loss calculation is relevant to an individual defendant’s culpability, and the departure for multiple loss causation is driven by the knowledge that, on occasion, the proxy will overstate culpability. In sentencing these defendants, the district court made explicit findings as to their relative culpability, rating the brothers Rostoff “at the high end of culpability,” Harris “somewhat lower,” Bonaiuto “somewhat below [Harris],” and “DiCologero below that.”12 The court then linked the degree of departure to the degree of culpability. Once a departure-justifying circumstance has been identified, and the sentencing court has determined to act upon it, a construct that varies the degree of departure based on relative culpability (as related to the actual ground for departure) seems eminently reasonable.13
E. Recapitulation.
We have made the pilgrimage that Rivera and Diaz-Villafane demand. Having done so, we find that the district court departed for an encouraged reason, permissible under the guidelines; that the departure-justifying circumstance is sufficiently record-rooted; and that the extent of the departures is within acceptable bounds (if barely). Consequently, we uphold the downward departures as to the defendants Harris, Bonaiuto, and DiCologero.
III. ROLE IN THE OFFENSE
The final leg of our journey brings us to the sentences imposed on the Rostoff brothers. Those defendants erect an immediate roadblock, asseverating that the district court’s downward departures eliminate any need to scrutinize the antecedent role-in-the-offense adjustments. Therefore, they urge us to vault directly to a departure analysis, ignoring possible errors in the court’s interim sentencing adjustments. We demur: following this course would put the cart before the horse.
We need not tarry, for the Rostoffs’ importuning impales itself on the horns of stare decisis. The reasonableness of a departure depends on its extent — and the extent of a departure cannot be measured unless and until a defendant’s sentencing range is established. Thus, “a decision to depart does not, as a general rule, render moot questions concerning the appropriateness of the calculations underbracing the district court’s computations of the GSR.” Emery, 991 F.2d at 910. The case at hand falls squarely within the Emery doctrine: each of the challenged role-in-the-offense adjustments “at least potentially, has more than academic effect on the actual sentence because the proportionality of the departure to the GSR is a salient factor to be considered in judging the departure’s reasonableness.” Id.
Having dismantled the Rostoffs’ roadblock, we turn to the challenged adjustments. The sentencing guidelines provide for elevating the OL of “an organizer or leader of a criminal activity that involved five or more participants or was otherwise extensive” by four levels, U.S.S.G. § 3Bl.l(a); elevating the OL of lieutenants — the “manager[s] or supervi-sores]” of such an activity — -by three levels, U.S.S.G. § 3Bl.l(b); and elevating the OL of those occupying leadership slots in smaller or less extensive criminal enterprises by two levels, U.S.S.G. § 3Bl.l(c). Here, the district court invoked subsection (c), and increased the OL of each Rostoff brother by two levels. The government contends that the court should have applied either subsection (a) or (b). We agree.
*413 A.What Transpired Below.
The disputed role-in-the-offense adjustments originated with the Probation Department. It recommended two-level enhancements under subsection (c) even though it acknowledged in the PSI Reports that the Rostoffs were “principal[s]” who “participated in the management and coordination of the scheme” and who “received a larger share of the proceeds of this conspiracy.” The government objected to the proposed adjustments, emphasizing the size and complexity of the plot. The Probation Department stood firm. Curiously, however, even while rejecting the objection, it conceded in an addendum to Steven Rostoffs PSI Report that the criminal activity was “extensive,” and that all five defendants had been “principal participants” in it.
The government renewed its objection before the district court, but to no avail; Judge Zobel accepted the Probation Department’s recommendations on this subject without making any independent findings. Accordingly, each brother received a two-level enhancement under subsection (c).
B.Standard of Review.
Role-in-the-offense determinations are innately fact-specific. The court of appeals must, therefore, pay careful heed to the sentencing judge’s views. See United States v. Ocasio, 914 F.2d 330, 333 (1st Cir. 1990). It follows that our standard of oversight is deferential: “absent mistake of law, we review such determinations only for clear error.” United States v. Dietz, 950 F.2d 50, 52 (1st Cir.1991). Questions of law engender de novo review. See United States v. Brewster, 1 F.3d 51, 54 (1st Cir.1993).
C.Analysis.
In ruling that subsection (c) applied, the district court necessarily found that the Ros-toffs were “organizer[s], leader[s], managers] or supervisor^]” of the criminal enterprise. U.S.S.G. § 3Bl.l(c). Neither side has challenged this finding. The question on appeal, then, is whether the defendants’ criminal activity “involved five or more participants or was otherwise extensive,” and, thus, fell outside the parameters of subsection (c).
The government’s assertion that the criminal activity involved at least five participants is ironclad. For one thing, the Probation Department’s finding to this effect is essentially unchallenged. For another thing, inasmuch as the jury found all five defendants guilty on the conspiracy count, the sentencing court was bound to conclude that the criminal activity involved no fewer than five participants. See United States v. Weston, 960 F.2d 212, 218 (1st Cir.1992) (explaining that under the guidelines “a guilty verdict, not set aside, binds the sentencing court to accept the facts necessarily implicit in the verdict”).
Despite the impeccable provenance of this fact, the brothers try an end run around it. They contend that U.S.S.G. § 3Bl.l(a)-(b) does not apply because, while they may have exercised leadership in a criminal enterprise that had at least five members, neither of them recruited, controlled, or directly supervised four other people.14 We need not dwell upon the correctness of the Rostoffs’ self-assessment, however, for their end run takes us on a fool’s errand.
Since the relevant language of subsections (a) and (b) is disjunctive, either extensiveness or numerosity is a sufficient predicate for a three- or four-level upward adjustment. See United States v. Hall, 996 F.2d 284, 287 (11th Cir.1993); Dietz, 950 F.2d at 53-54. In this instance, a careful review of the record leaves no room to doubt the extensiveness of the criminal enterprise. Thus, we need not inquire into the attributes that might — or might not — be essential if the enhancement depended upon a finding of nu-merosity.15
*414Unlike numerosity, extensiveness does not depend upon a finding that a criminal activity embraced no fewer than five criminally responsible participants, see United States v. Melendez, 41 F.3d 797, 800 (2d Cir.1994); Dietz, 950 F.2d at 53-54-, much less a finding that the activity included four or more persons under the defendant’s direct control. Rather, a determination that a criminal activity is “extensive” within the meaning of section 3B1.1 derives from “the totality of the circumstances, including not only the number of participants but also the width, breadth, scope, complexity, and duration of the scheme.” Dietz, 950 F.2d at 53.
Here, the conspiracy lasted for over three years, encompassed a bare minimum of 140 fraudulent loans, consumed millions of dollars, affected many lives, and involved a legion of people beyond the five named defendants. On this record, we are compelled to conclude that the defendants’ criminal activities satisfy the extensiveness standard that is built into U.S.S.G. § 3Bl.l(a)-(b). Consequently, the two-level enhancement cannot stand: if the district court impliedly held that the defendants’ criminal activity was not extensive, it committed clear error, and if the court applied section 3B 1.1(c) notwithstanding the extensiveness of the criminal activity, it misapprehended the law. Either way, the court’s crafting of the Rostoffs’ adjusted offense levels undervalued their respective roles in the offense, requiring resen-tencing.16
IV. CONCLUSION
We need go no further. For the reasons stated, we affirm the convictions of all defendants, and affirm the sentences meted out to the defendants Harris, Bonaiuto, and DiCo-logero. However, we vacate the sentences imposed on the defendants David and Steven Rostoff, and remand their cases for resen-tencing in light of the need for altered role-in-the-offense determinations.
Affirmed in part, vacated in part, and remanded for further proceedings consistent with this opinion.
. We note two related facts. First, after investigators discovered the fraud, Bonaiuto falsely asserted that he had queried borrowers about the deposits shown on the settlement sheets, and that they had assured him that they had made the indicated down payments. Second, Bonaiuto also acted as closing attorney for the bank in connection with his own purchase of two condominium units from the Rostoff group. On each occasion, he submitted a settlement sheet to the bank showing that he had tendered a 10% down payment when, in fact, he had made no down payment at all.
. One vignette is particularly telling. On July 23, 1987, DiCologero's husband closed a mortgage loan at the bank in order to finance his purchase of a condominium from the Rostoff group. The settlement statement falsely indicated that a $7,700 deposit had been made when, in fact, DiCologero's husband had purchased the condominium with no cash down payment (advancing only $1,663.40 in closing costs). The record shows that DiCologero shepherded the loan through the bank's approval process.
. The jury convicted the defendants on a count that charged a conspiracy beginning in 1985 and continuing into 1989. It is well established that the sentencing guidelines apply to offenses that straddle the effective date of the guidelines (November 1, 1987). See United States v. David, 940 F.2d 722, 739 (1st Cir.), cert. denied, 502 U.S. 989, 112 S.Ct. 605, 116 L.Ed.2d 628 (1991). Even in such cases, however, the guidelines in effect at the time of sentencing, not those in effect at the tag end of the offense, ordinarily control at sentencing, except where ex post facto concerns loom. Cf., e.g., United States v. Harotunian, 920 F.2d 1040, 1041-42 (1st Cir.1990). The district court, invoking this exception, applied the November 1987 version of the guidelines. No party questioned that choice below, and no party asks us to revisit it on appeal. Since we follow the district court's lead, all references herein are to the November 1987 edition of the guidelines unless otherwise specifically indicated.
. Inasmuch as the Rostoffs must be resentenced for other reasons, see infra Part III, we limit our departure inquiry to the sentences imposed upon Harris, DiCologero, and Bonaiuto, respectively.
. The November 1990 version of application note 11 is identical to the 1987 version and, thus, controls in this case. See supra note 3.
. The new note reads in pertinent part:
In cases in which the loss ... does not fully capture the harmfulness and seriousness of the conduct, an upward departure may be warranted.... In a few instances, the loss ... may overstate the seriousness of the offense. This may occur, for example, where a defendant attempted to negotiate an instrument that was so obviously fraudulent that no one would seriously consider honoring it.
U.S.S.G. § 2F1.1 comment., n. 10 (Nov.1991).
. The government also suggests, in what it bills as a separate argument, that the district court improperly relied on the conduct of the bank and of the buyers as a basis for departing. At bottom, however, this suggestion is predicated on the government’s assertion that it is improper to focus on any causes of the loss apart from the conduct of the defendants. As we have pointed out, such a position is inconsistent with both the unambiguous language of the original commentary that accompanied section 2F1.1 and the clear import of existing circuit precedent. Hence, the government's "separate” suggestion adds nothing to its flagship argument.
. At any rate, the record buttresses the district court's conclusions. The evidence establishes that bank officials approved myriad loans, total-ling millions of dollars, with an abandon commonly associated with drunken sailors. In the bargain, senior management routinely authorized loans that exceeded the bank's LTV ratio, backdated documents, and acted, to use the government's phrase, in an “incredibly negligent” fashion. The evidence also shows that many of the purchasers were sophisticated investors who, enthralled by gimmicks like the phantom down-payment concept, bought multiple properties. As sophisticated investors surely should know, projected profits that look too good to be true often are — and often signify the presence of great financial hazards. Finally, an economist's affidavit, introduced at sentencing, graphically illustrates the extent to which the bottom fell out of the condominium market.
. The following chart illustrates the degrees of departure:
Defendant GSR Incarcerative Sentence
J. Harris 27-33 months 9 months
P. Bonaiuto 33-41 months 0
D. DiCologero 27-33 months 0
Relatedly, the court placed Bonaiuto on two years probation, six months of which was to be served in a community treatment center, and sentenced DiCologero to a two-year term of straight probation.
. In general, departures can be classified as either "guided” or "unguided.” As the label implies, a guided departure is one in which a "guideline or related commentary suggests that under [the] particular circumstances the departure should be calibrated by a particular analogy to the sentencing grid.” Bruce M. Selya & Matthew R. Kipp, An Examination of Emerging Departure Jurisprudence Under the Federal Sentencing Guidelines, 67 Notre Dame L.Rev. 1, 12 (1991). In contrast, an unguided departure, although it may be based on grounds specifically encouraged or identified in the guidelines, is not constrained by a specification of the means through which the sentencing court must calculate the departure's magnitude. See id. We restrict our discussion today to unguided departures, because former application note 11 to section 2F1.1, as it appeared in 1987, offered no definitive directions for determining the extent of downward departures based on multiple loss causation.
. Our concurring brother misapprehends this point. Since reasonableness is necessarily a function of what the sentencing court knows, depriving the court of the prosecutor's judgments about the extent of an anticipated departure limits the court's knowledge and, thus, affects the reasonableness of its ensuing determination. Contrary to Judge Stahl's assumption, the unhelpful prosecutor does not “waive'', anything; *411he simply makes his post hoc complaint less convincing.
By like token, we do not believe that we are encouraging "empty exercise[s]." Post at 415. We agree that the prosecutor who, as our concurring brother suggests, "recommend[s] a downward departure of one week,” does not assist the sentencing court. Id. We disagree, however, that such a ruse would improve the prosecution's position or help to alter the calculus of reasonableness. It should go without saying that, just as we expect lawyers who suggest jury instructions to base them on existing law or good faith arguments for new law, so do we expect the government to be candid and forthcoming in commenting upon the reasonableness of an anticipated departure.
. In a colloquy with the court, the prosecutor ranked the defendants in order of perceived culpability, listing David Rostoff as the most culpable, Steven Rostoff second, and Harris third. Bonaiuto and DiCologero brought up the rear. For the most part, the sentences imposed by Judge Zobel coincide with this ranking. This parallelism makes it all the more difficult for the government to maintain that the judge’s method was madness.
. Of course, relative culpability alone is not a reason to depart. See United States v. Wogan, 938 F.2d 1446, 1448 (1st Cir.), cert. denied, 502 U.S. 969, 112 S.Ct. 441, 116 L.Ed.2d 460 (1991). If, however, a valid departure-justifying circumstance is present, and the sentencing court acts on it, relative culpability appropriately can influence the degree of departure.
. The operative number of other persons is four rather than five, since the defendant himself must be counted as a participant, see United States v. Tejada-Beltran, 50 F.3d 105, 109 n. 9 (1st Cir. 1995), and the defendant presumably is under his own control.
. , Some courts have held that, when the applicability of § 3Bl.l(a) depends upon numerosity rather than extensiveness, the defendant must be shown personally to have controlled no fewer than four other participants. See, e.g., United States v. Carson, 9 F.3d 576, 584 (7th Cir.1993) *414(stating that to warrant invoking subsection (a), the defendant must have had "some control, direct or indirect, over at least four other participants in the offense”), cert. denied, -U.S. -, 115 S.Ct. 135, 130 L.Ed.2d 77 (1994); United States v. Reid, 911 F.2d 1456, 1465 n. 8 (10th Cir.1990) (same), cert. denied, 498 U.S. 1097, 111 S.Ct. 990, 112 L.Ed.2d 1074 (1991). Other circuits take a different position. See, e.g., United States v. Dota, 33 F.3d 1179, 1189 (9th Cir.1994), cert. denied, -U.S. -, 115 S.Ct. 1432, 131 L.Ed.2d 313 (1995). Both the validity and the permutations of this interpretation of § 3Bl.l(a) are open questions in this circuit.
. The government maintains that David Ros-toff's offense level should be enhanced by four levels pursuant to U.S.S.G. § 3Bl.l(a) and Steven Rostoffs offense level should be enhanced by three levels pursuant to U.S.S.G. § 3B 1.1(b). We take no view of these particulars, leaving the resolution of such interstitial questions to the district court.