Here's one more entry in the ongoing investigation of what's up with the tracks that are not matched back to generated particles in our efficiency calculation.  The previous entry in this investigation can be found here:

https://drupal.star.bnl.gov/STAR/blog/dmawxc/update-06212019-run-9-embedding-tower-matched-vs-not-tower-matched-tracks-are-not-matche

This time around, I did the efficiency calculation but only used reconstructed recoil tracks that were matched to towers for the "detector level."  I made no such stipulation regarding the towers at the particle level.  Here's the result:

Surprisingly, the efficiency is more well-behaved than our normal efficiency (meaning that it more closely follows the typical '1 - exp[-x]' form, there's no "hump" at 'pTtrk = 2'). But it's also substantially lower.  The asymptotic value could probably be pushed up by only considering particle-level tracks that we would anticipate to be matched to a tower.

Here's the same calculation, but using the reconstructed recoil tracks matched to towers that were not successfully matched to generated particles as the "detector level:"

No surprise here: it's just gibberish.  And here are the two "efficiencies" compared to each other: