Based on the plots here:

http://drupal.star.bnl.gov/STAR/blog-entry/aliceb/2009/feb/02/emc-meeting-2-february-2009

it was suggested that I remove all ped dependence on the database from the maker.  I changed StEmcADCtoEMaker as follows:

 StEmcADCtoEMaker *adc = new StEmcADCtoEMaker();
 adc->saveAllStEvent(true);//

New distributions for the bprs and smd at the nominal time point (run 8330094) are attached.  Things do not look significantly different for the preshower and while there is no longer a strange cutoff for the shower max, there is still not much hope of fitting a slope.

Furthermore, as can be seen in timings.pdf, the spectrums don't look significantly different at different time points.  In this plot a show, for a single bsmde channel, the adc spectrum at most extreme time points as compared to the nominal, it the three different colors.  FYI black is a time delay of -150 ns, dark blue is the nominal and light blue is a delay of +150ns.

At least one problem is being statistics limited.  The only option I have now to try to improve the statistics is to try to add multiple channels together rather than doing all channels separately and then taking the average for each crate.  This might not be the right thing to do.