I've had a chance to incorporate the st_upsilon production into my check on the BEMC tower calibration.  There are now enough statistics to look at individual towers (see electrontowers.pdf).  Forthcoming will be a list of towers that need their calibration coefficients corrected.  Below is a plot of (E - p)/p for every electron tower.  Red lines are at 10% and the green line is a fit to a constant.

My previous post showed a plot of E/p for each η ring.  With additional statistics, the plot now looks like:

There was some question about whether this plot indicated a change needed to be made to the calibration coefficients in the outermost bins in eta.  A look at the E/p spectra for those rings (see attached electronring.pdf) shows that the integrated curves can't be used to make new coefficients.  This analysis was done in 2006 (see attached electron_normalization_2006_nov07.pdf courtesy of Adam) for pp to correct for a shift between the mip calibration and the electron calibration.  A look at the neutral pion mass peak for 2006 in the outermost eta bins (see attached etabinned.pdf and etabinned2.pdf courtesy of Alan, bins 20 and 40) tells us that the updates to the calibration coefficients was pretty close to right on.

A look at a pure sample of electrons from the Upsilon production by observed the opposite behavior in a plot of p/E.  I successfully recreated the plot using their cuts (0.5 < p/E < 1.4, p > 3) below:

This indicates the outer eta rings calibration coefficients are high rather than low.  This effect is clearly a result of the p/E cuts.  If you look at the ring histograms in electronring.pdf, you'll see that cutting on 0.7<E/p<2.0 clearly shifts the mean of each plot upwards, especially in the outer eta rings.  This only strengthens the notion that we should avoid drawing conclusions about the outer eta based on ring integrated plots.

Additional discussion on this topic: http://www.star.bnl.gov/HyperNews-star/get/emc2/2558.html