Hi all,

 as I pointed out many times in the last days, the new code to perform
semiexclusive reconstruction on the skims picks up dirty modes (but
the yields are also larger).

 In our new sample we can find a lot of modes that originally belong to
the superblock4 (purity of the mode < 10%).

 Then a reoptimization and a reordering of the modes according to the
single mode purity was mandatory.

 In web page:

http://www.slac.stanford.edu/BFROOT/www/Physics/Analysis/AWG/InclusiveSL/islrecoil/QA_08312002.html

 I added a table of the modes ordered according to their purity (in red).
In the last three columns you can find:

 1) integrated plot (adding up all modes after the ordering)
 2) integrated plot after the lepton cut
 3) integraten plot after all cut(except the kaon cut)

 Then I performed the optimization.
 The idea is that the cut on the purity has to:

 - maximize the stat significance for our analysis
 - avoid biases in the mes fit for the two level of cuts: lepton cuts, all
   cuts


 In the web page I add the plot (per mode) for the S/sqrt(S+B) and purity
vs total yield for

 - no cut
 - lepton cut
 - all cuts (no kaon cut)
 - all cuts + kaon cut + mx < 1.55


 Then the second optimization criterion (no bias) was studied.

 Since the quantity that has to be unbias is Nallcuts/Nleptoncut I plotted
the yield after all cuts versus the one versus the yield after the lepton
cut and the ratio of . The point in which the ratio changes
dramatically indicates the beginning of a bias and this is the right
place to put a cut.


 From all there inputs I got these optimal points (all values of purity
are the new recalculated ones).

 - D*   ipur > 14%
 - Dc   ipur > 6.5%
 - D*0  ipur > 5%
 - D0   ipur > 6%


 Comments:

 - The slope in all the ratio(allcuts/lepcuts) plots is not good and it
should be investigated. We should understand whether it is due either to
a different efficiency or to a bias in the fit.

 - With these cuts we will have ~10000 events after all cuts in 85fb-1 (a
factor 2.2 the one in the old production for 51fb-1).

 - The statistical error in Vub (taking into account the decrease in
purity) will decrease of a factor 35%.



   Daniele