During last weeks jet meeting, Tai said that he found a number of towers included in the final 2006 A_LL result which consistently fired on background events.  He claimed that if we remove these towers that the measured A_LL increases (less dilution).  We were curious... so shown below is the tower Id of the highest E_T tower in the jet (for the range 7.0<E_T<13.2) for the 2006 data sample.  This is using the OR-sum trigger condition and after we have applied the usual R_T<=0.92 cut.

Now the same figure using a tighter R_T<=0.80 cut:

And finally after removing the towers 622 towers that Tai pointed us to:

I don't have jet trees with these towers excluded from jet-finding.  What I can do though is remove all jets with this tower as the highest E_T tower in the jet.  That's what is done in the figure above.

Now, the interesting question though is the effect of these towers on A_LL.  Shown below is A_LL using all the towers with the standard 0.92 cut (black points), A_LL after cutting on jet phi |phi|>0.7 (blue triangles), and A_LL after removing all of Tai's bad towers (red triangles).