Jet constituent vertex information in runPerformance.py

[Brainz22]

FastPUPPI/NtupleProducer/python/runPerformanceNTuple.py

Line 333 in ac37b69

setattr(process.l1pfjetTable.moreVariables, "dau%d_%s" % (i,"vz"), cms.string("? numberOfDaughters() > %d ? daughter(%d).%s : -1" % (i,i,"vertex.Z")))

I find that daughter.vertex.Z is not consistent with pfTrack()->vz() used in L1PFCompare.cc#L348, which is consistent with this CMSSW line (without the primary vertex subtraction). Consequently, daughter.vertex.X is not consistent with pfTrack()->vx() and daughter.vertex.Y is not consistent with pfTrack()->vy(). These feature differences lead to score miscorrelations between qkeras model, HLS (from hls4ml) model, and the CMSSW TOoLLiPProducer emulation.