[Recorded on 05.31.2019] Following up on the previous post:

https://drupal.star.bnl.gov/STAR/blog/dmawxc/update-05272019-run-9-embedding-all-vs-matched-reconstructed-tracks

As the title implies, we found the bug(s). The issue was two-fold:

1. I was using the wrong variable for selecting recoil tracks in the "all reconstructed tracks" category. It was a typo in my code, and I was essentially cutting reconstructed tracks at random (but only in the "all reconstructed tracks" category). This explains why the integral of the "all reconstructed tracks" spectrum was significantly smaller than the "matched reconstructed tracks" spectrum even though the "matched" spectrum should be a subset of the "all" spectrum.
2. In the code, I require a trigger at the particle-level and at the detector-level. However, I make no stipulation that these are the same particle. So -- since we're looking at dijet events -- there are situations where the detector-level trigger could be part of the recoil jet corresponding to the particle-level trigger. Hence, we see a dip in the efficiency in the pT range corresponding to our trigger selection.

There are a number of fixes for (2), but after discussing we agreed it was enough to require a particle-level trigger and nothing else. Here are plots illustrating the differences between all and matched reconstructed tracks after the fixes:

We were surprised by just how much additional tracks there are on top of the matched tracks, so we're going to look into those (which should just be pileup from the zero-bias pp collisions which the Pythia events are embedding into) before continuing on with the analysis proper.