Hi all, I produced the plots that show the resolution as a function of mm2 both for mxhad and mxhadfit for the new selection. b->clnu http://www.slac.stanford.edu/~daniele/vub/compvcbmm2.eps b->ulnu http://www.slac.stanford.edu/~daniele/vub/compvubmm2.eps They are produced on the cocktail MC. The cut applied are - one lepton - pcms > 1 - |mm2| < 1 I did the same as a function of the true Mx b->clnu http://www.slac.stanford.edu/~daniele/vub/compvcb.eps b->ulnu http://www.slac.stanford.edu/~daniele/vub/compvb.eps If I compare with the ones in the BAD (page 24-25) I notice that: 1) now the Mxhad resolution is better 2) the benifit from the kinematic fit is lower 3) a bias is present for b->clnu (~100 MeV). Before (BAD) the mean was always unbiases Do we have to train again the kinematic fit with the new selection track and neutral selection? BTW, Oliver, may you confirm that I applied the same cuts used to produce the plots in the BAD and that you are plotting mean and RMS and not the result of a gaussian fit? Daniele