Just as a stupid cross-check I run the MC fit on Run1+2 and Run4 separately. I remind you that I am taking the same events as both data and MC, and not applying any reweighting at all. Generation value: 0.0198 Run1+2 fit: 0.0227747 +- 0.00140784(stat) +- 0.00117712(MC stat) Run3 fit: 0.0201367 +- 0.00170562(stat) +- 0.00155887(MC stat) Run4 fit: 0.020348 +- 0.000950046(stat) +- 0.000755165(MC stat) Run1-4 fit: 0.0210372 +- 0.00072623(stat) +- 0.000590931(MC stat) It seems that the MC fit is OK for Run4, Run3 and not for Run1+2, and consequently for Run1-4! Results in http://www.slac.stanford.edu/~bozzi/scra/IbuMCmXSunfixfitR12/ (Run1+2) http://www.slac.stanford.edu/~bozzi/scra/IbuMCmXSunfixfitR3/ (Run3) http://www.slac.stanford.edu/~bozzi/scra/IbuMCmXSunfixfitR4/ (Run4) http://www.slac.stanford.edu/~bozzi/scra/IbuMCmXSunfixfit/ (Run1-4) The discrepancy is mainly due to the number of fitted Vub events in the 1st bin: Run1+2: gen. 591, fit 687.972 +- 30.0696(stat MC) +- 42.5277(err fit) Run3: gen. 289, fit 303.269 +- 19.4599(stat MC) +- 25.6875(err fit) Run4: gen. 1471, fit 1515.57 +- 42.7063(stat MC) +- 70.7618(err fit) Run1-4: gen. 2352, fit 2521.15 +- 55.9467(stat MC) +- 87.033(err fit) What might be wrong? The only reason which might explain this behaviour is (again...) that S/peakingBG might be different for different run periods whereas I am using the values computed on the entire sample, fitted with a 1st order polynomial. But, are there any reasons why S/peakingBG should be substantially different for different run periods? I quickly investigated this by computing the S/peakingBG which comes out from a (argus+ccb+signal) fit on the entire mx sample on charged B events, getting: S/peakingBG: Run1+2: 3.85 +/- 1.38 Run3: 5.09 +/- 1.32 Run4: 3.95 +/- 0.35 Run1-4: 4.00 +/- 0.28 which are compatible within errors, and in any case going in the wrong direction for what the yields are concerned. Anything else? Concezio.