Hoi Alessio,

very nice talk! Here are just a few nit-picking comments: 

p2: "model dep." not due to  mb, a parameters, but rather due to shape
    function parametrization. The parameters  per se are HQET "theory"
    parameters,   not  a  model.    But  the   specific  form   of  SF
    parametrization in deFazio/Neubert is a model. Which is THE reason
    why we want to avoid a dependency on the SF.

p4: I  would flip the x-y  axes for the "correlation"  plot, then it's
    the  same as  other people  use, e.g.   in the  canonical  plot of
    Bauer/Ligeti (Fig  1 of ph/0107074).   I think in general  this is
    called a "Dalitz" plot, not a "correlation" plot.

p5: I do not think that we need  to go as high as 10 GeV2. In fact, IF
    we  want to  put out  a result  as soon  as possible  with minimal
    changes, I would  argue that the q2 cut  should be relatively low,
    with higher mX cuts than what  we have done so far. See e.g. table
    I in ph/0107074. 

    Why 10Gev2? 

    I have to admit that I  have not yet clicked through all the links
    in your webpage,  I don't know whether you have  done this. Ed has
    produced scans in the (mX, Q2) plane.
   
p6: I think that this should be  a 2-d scan in (mX,q2), not just a 1-d
    scan. You must look at different combinations.
 
    Not only "sys" reduction, but "sys^2 + theo^2". We will not gain
    in "sys" probably. 

p7: Naive  question: Why does the  efficiency flatten at  mX = 1.8GeV?
    Applying a  q2 >  10 GeV2  should leave signal  rate to  higher mX
    (both from your plot on page 4 and from ph/0107074).

    Again, I think that it should be a 2-d scan of the error not only
    vs. q2, but in the full (q2,mX) plane. 

p8: By now I think I was confusd and what you call "sys" is actually
    "theo";  OK...

p9: I would  propose that we very seriously  consider switching to the
    non-resonant  as  "default"  signal  MC.   Of course  we  need  to
    understand the behavior of the error on the hybrid MC.

p10 I would feel (very) uncomfortable to extract mb from a spectrum
    that shows a disagreement between the model and data as shown in
    the bottom left two plots. 

    Before we  do that we should  be sure that  e.g. changing neutrals
    selection and tracking  has a very small effect  on this. There is
    work  ongoing by  Jonathan Coleman  which  we should  look at.   I
    consider this  one of the  more important improvements (on  a time
    scale longer than this summer) for the incl b2u analyses.

p12 Again, I personally would not claim that we can extract mb from
    the mX distribution to 80MeV. 

p18 "1." I would maybe say that we need to understand the hybrid MC
    "For this summer 2." I would say we need to validate the spectra
    other than mX given from the fit. 

That's all for now.

Cheers,
--U.