In order to further check the neutral selection (which was estimated on data), I checked the behaviour of the two chosen variables on MC. I produced the usual recoil vs reco plots for neutral candidates that are or are not matched with the MC truth: LAT for candidates mathed with MC truth http://www.slac.stanford.edu/~rfaccini/phys/vub/aug02/Lmom_trueMC.ps LAT for candidates NOT mathed with MC truth http://www.slac.stanford.edu/~rfaccini/phys/vub/aug02/Lmom_falseMC.ps both the lower and the upper cut are reasonable also on MC S9S25 for candidates mathed with MC truth http://www.slac.stanford.edu/~rfaccini/phys/vub/aug02/S9S25_trueMC.ps S9S25 for candidates NOT mathed with MC truth http://www.slac.stanford.edu/~rfaccini/phys/vub/aug02/S9S25_falseMC.ps there are clearly no true photons below 0.9, so that the cut is not dangerous. The eddect of the cut is not striking, but there are some tails at low S9S25 in the 'false' sample. The most remarkable thing is that there are a lot of fake photons at S9S25==1, which means very narrow clusters. These are already cut (at least partially) by the LMom>0.05 cut ciao ric