It was pointed out during my presentation to JetCorr on May 22nd, 2018 that while the Run 9 embedding and Run 9 data recoil jet (R = 0.3, charged) 'pT'  distributions fall on top of each other, the Run 12 embedding recoil jet (R = 0.3, charged as well) distribution was consistently above data.  For reference, the comparison between the Run 9 embedding and data jets can be found on slide 11 of this presentation:

https://drupal.star.bnl.gov/STAR/system/files/JetCorrUpdate_Anderson.May22.v1.pdf

And the comparison between the Run 12 embedding and Run 9 data jets can be found on slide 6 of this presentation:

https://drupal.star.bnl.gov/STAR/system/files/FullJetReconstructionPP_Anderson.CollaborationMeetingJan25.v1.pdf

The difference between the Run 9 and Run 12 embedding distributions is the trigger: the Run 9 jets are from charged hadron-triggered events while the Run 12 jets are pi0-triggered events.  So there is an order of magnitude less statistics in the Run 12 sample.  There are 15093 and 12291 charged hadron triggers in the RFF and FF configurations of the Run 9 embedding, and only 2408 pi0 triggers in the Run 12 embedding.

If you plot all 4 distributions (data, Run 9 RFF, Run 9 FF, and Run 12) on top of each other, they are consistent within error bars.  Note that the Run 12 per-trigger yields are a factor of 2 higher than in the original presentation above.  This is to compensate for the binning: the data and Run 9 distributions were originally binned with 'dPt' = 0.5 GeV/c, but were re-binned to 'dPt' = 1 GeV/c.  Since I didn't normalize for bin-size, I just scaled the Run 12 distribution (which was binned with 'dPt' = 1 GeV/c) by 2.