I am trying to figure out why the trigger simulator only returns as 'good' a fraction of the events returned as 'good' by the online L2-gamma trigger.  Let me state that a different way...

The above plot is the trigger summary for run 7136073.  For each event in the run I ask "does this event satisfy the online L2gamma trigger (137611)"  If it does I fill the first column.  Then I ask "does this event satisfy 137611 at the simulator level?"  If it does I fill the second column.  If it satisfies both, I fill the third column.  As you can see, there is a discrepancy between the number of online and simulated triggers to the tune of about 30%.  Why?

Next I ask, for each event that passed the online trigger but failed the simulated trigger, which subsystem passed? 

the four columns represent bbc (mb), bemc (BHTTP1), L2 (L2gamma), and total number of failed events.  The vast majority of those events that failed did so at the BHTTP1 level.  Finally, I dumped this information into a text file (only for about half the events for size considerations.)  The text file contains, for each event that fails the simulated trigger, event number and trigger decision for each subsystem.  that can be found here.