Hi all, since we are finalizing our measurement and we are taking care of all the higher order effects, I studied more in detail the ratio of the breco efficiency. Few days ago I noticed strange behaviours of the breco efficiencies depending on the recoil multiplicity. The first observation was on the efficiency for the different charm components. If I look at the MC truth in the recoil and I ask for a well reconstructed B, I get these numbers (for B0's): Decay Nevents BR/BR(totSL) BR/BR(tot)MC breco effect B->Dlnu 43365.9 +- 366.797 0.194% 0.186% +4% B->D*lnu 124123 +- 628.731 0.556% 0.496% +12% B->D1lnu 12056.5 +- 221.803 0.054% 0.050% +8% B->D2lnu 7579.53 +- 174.735 0.034% 0.033% +3% B->D**lnu 36350 +- 384.04 0.163% 0.236% -44% B->Xclnu(total) 223364 +- 870.031 As you see the effect of the B reconstruction seems to be dependent on the semileptonicmode on the recoil. Two explanations are possible: 1) the MC BR are wrong. The BRs have been calcualted at generator level. 2) breco is dependent on the multiplicity of the recoil Possibility 1) has to be excluded if the computation at generator level was ok. If we think that the possibility 2) is feasible we should correct for this effect since b->ulnu have a lower multiplicity than b->clnu, thus introducing a bias in the ratio of BR. I remind you that in the BAD we just say that we verified that the ratio of the breco efficiencies (eff(b->ulnu)/eff(b->clnu)) is compatible with 1 within the stat error and we fixed to 1 that ratio, since no dependency on the recoil decay mode is expected. The effect observed on the D**lnu raises doubt about this assuption. I recalculated this ratio on the full enlarged generic MC sample running B0 reco on B0 MC and the same for Bch in order to reduce background and rely on the fitted yields. These are the results: B0 ratio(epsbreco) = 1.069 +- 0.053 Bch ratio(epsbreco) = 1.054 +- 0.044 and the weighted average is ratio(epsbreco) = 1.06 +- 0.04 ------------------------------- As a crosscheck I also recalculated the famous pstar factor (i.e. the ratio of the lepton cut efficiencies) using all the available MC (generic + cocktail) with a cut on the integrated purity and now I get ratio(epssl) = 1.12 +- 0.02 (we were using 1.09) ---------------------------- My proposal is then the following: - take the ratio of the breco efficiency we get from the generic MC, that is 1.06+-0.04 and rescale accordingly the result. Quote the difference from 1 as a breco systematics (6%) - use the new pstar factor 1.09->1.12(and its stat error will go in the MC statistics systematics). The effects of these changes will be: - the central value will go down by 9% (0.0196->0.0180) - the total experimental systematic uncertainty will increase by 10% (12%->14%). The extra contribution will be called a breco systematics Let me know what you think about it. Daniele