In my last post I showed the ratio of BHT3coin and dijet triggers, revealing a discepancy in the middle of the run.  The plot below shows the BHT3coin/BHT3 ratio over the same span.  We see a systematic shift between the early runs and the later runs.  

I've had to retool the macro a bit to run both luminosity measurements simultaneously, but when that's done I can collect all the values I use for measuring the original non-coincidence luminosity at the same time as the coincidence luminosity.  I'm hoping this will make it clear what's causing the earlier discrepancy.

On the plus side, this suggests that in general both methods of measuring the integrated luminosity give similar values, save that BHT3coin has much smaller error.

Obviously, the area where BHT3 !~ BHT3coin is not the same as where dijet !~ BHT3coin.