At Matt Walker's request, I re-ran the code that computes the Z invariant mass from the 500GeV L2W data, using 'ofl' gains instead of 'sim' that had been used before.  The result is a fairly large shift in the weighted mean, but both are still more or less consistent with the expected value.

with 'sim' gains:

when we change the gains, we get:

There is a slight difference in the shape of the two plots, which may be due to slightly different performance of the cuts.  The number of jobs that failed has also changed, so the different shapes and numbers may be due to that, instead.

In the following plot, these two values are compared to the pdg Z mass (red line):

We see a difference of .88*sigma for the 'sim' flavor, and 1.76*sigma for the 'ofl' flavor.  Comparing just the two flavors against each other, there is a shift of ~4.2% in the position of the peak.