I haven't resolved a sharp discrepancy between data and MC shape after trigger simulator is applied.  The difference is most visible on page three of the attached pdf, which shows the number of events as a function of E_T for a small dataset and also for Mike's endcap photons+qcd.