Hello Osamu, 
Just realized I should have replied to your e-mail, but forgot: 

> During the meeting today, we were aware that the WCM (or specific 
> luminosity dist of polarimeter) is wrong for yellow. And we thought
> that I(Osamu) had made mistakes to use the blue WCM in the yellow analysis.
> 
> But it turned out that the plot I showed was rael, it means "the yellow
> had the bad 5 bunches as well".
> Please take a look at follwing two kinds of plots and confirm this fact.


What we observe in STAR: 
- specific luminosity: there is a group of a few bunches:
( bunch crossing 36-40, i.e. 
Yellow bunches 36-40 collided with Blue bunches 16-20)
which has specific luminosity smaller or bigger than 
for the other bunch crossings. 

For example fill 2251:
www.physics.ucla.edu/~joanna/Fill2251/Run3018025_WCM.ps  

to be compared with your plots for Fill 2251, run 755 (Blue) 757 (Yellow):
http://ccjsun.riken.go.jp/~jinnai3/polarimeter/bunch_sel/Specific/index.html
http://spin.riken.bnl.gov/exp/pcpol/2001/Bunch_web/index.html


Both STAR and CNI data (specific luminosity) are consistent. 


The problem with bunch crossings 36-40 I interpreted as originating from 
Blue beam ( two groups of weak bunches from intensity plot). 
Yellow looked perfectly "normal" to me, i.e. intensity distribution 
was flat.
 
>From your(Osamu's) plot however (yellow and blue specific luminosities:
specific luminosity for yellow has groups of bad bunches although 
intensity distribution is flat, specific luminosity for Blue is flat 
although intensity has 2 groups of weak bunches) one can
see, that problematic are Yellow bunches. 
 In STAR (and Phenix) "bad" Yellow bunches collide with low intensity
Blue bunches or abort gap ( bunches became bad due to a collision with low
intensity bunches? ).

So, removing 2 groups of Yellow and 2 groups of Blue bunches
results in removing only one group of bunch crossings 
in STAR (and PHENIX).

  Regards, Joanna