So, my investigation into the Run 9 efficiency continues, and it's only getting more confusing.  Previous studies can be found here:

https://drupal.star.bnl.gov/STAR/blog/dmawxc/update-02192019-run-9-pp-track-efficiency-vs-ettrg

My original intent was to compare the detector-level Pythia8 and Embedding track pT distributions without smearing (ie. after applying the efficiency), take the ratio, and use that ratio to empirically calculate the "correct" efficiency. This depends on a few assumptions:

1. The particle-level Pythia8 and Embedding distributions agree if not are identical
2. The detector-level Embedding distribution describes the data well

Do those bear out? Here are the detector-level distributions compared to data:

And their ratios:

So assumption (2) is bogus. But what about (1)? Here are the particle-level distributions (plotted against data for comparison):

So (1) is also completely unfounded! For completeness, here is the ratio of the particle-level Pythia and Embedding distributions:

And also here's the extracted efficiency for the above embedding distributions:

Now I'm back to square one. I don't know what the issue is, and I don't which way to proceed. I'm probably doing something profoundly dumb, but I'm not sure what... The safest route is probably to re-do the efficiency calculation from scratch.  However, it might also be that the efficiency is perfectly fine and that issue is at the jet-finding level; ie. I'm just doing something different with the jet-finding in Pythia vs. data and embedding...  See this post, for instance:

https://drupal.star.bnl.gov/STAR/blog/dmawxc/update-01072019-run-9-pp-detector-level-pythia-vs-embedding-vs-data-jet-spectra

But, again, I'm not sure...