It has been suggested that the difference between the missing energy and the missing momentum, e-p, should have the same rejection power as (mm2=(e-p)(e+p)) but be more directly related to the measured quantities (and therefore more likely to be gaussian-distributed). I have therefore investigated this quantity. I have first checked the correlation between e-p and mm2 on background (top) and signal (bottom) MC: http://www.slac.stanford.edu/~rfaccini/phys/vub/eminusp/ximiss.eps note that there is a sign correlation between the two quantities, but it is not fully true because e+p can be negative, because of resolution effects, in a small fraction of cases. The overall correlation is not small. I then looked at the rejection power, comparing the signal (dots) with the background (histo) after all other cuts (removing the one on mm2) http://www.slac.stanford.edu/~rfaccini/phys/vub/eminusp/comparison7eminuspallcuts-sig.ps The rejection power is clear, I have arbitrarily picked up a cut at e-p< 300 MeV to check the performances (see later). I also looked at the data-MC agreement of this quantity: http://www.slac.stanford.edu/~rfaccini/phys/vub/eminusp/comparison7eminuspallcuts.ps I would say that all effects visible in the mm2 distribution are also present here, in particular the disagreement in the low tail is still present. Finally I applied the above mentioned cut and looked at the mxhadfit distribution for all candidates http://www.slac.stanford.edu/~rfaccini/phys/vub/eminusp/comparison7mxhadfitallcuts.ps and for b0 alone http://www.slac.stanford.edu/~rfaccini/phys/vub/eminusp/comparison7mxhadfitleptoncuts-b0.ps All effects we are worried about are still there. In conclusion this quantity seems to be highly correlated with the one we already use and switching to it seems not to bring advantages ciao ric