A detailed comparison between gen2 and gen3 middleware data products up to and including Single Frame Processing was made on
DM-28858. It was revealed there that there are small differences in the sources that actually get used in the photometric and astrometric calibrations of order a-few-to-a-dozen per CCD. This leads to small differences in the WCS solutions resulting in slightly different CCD pixel scales (up to ~0.08 mas), coordinate offsets (up to ~3.4 mas), and small differences in some photoCalibs (differences up to: calibrationMean: 2e-03 calibrationErr: 6e-07 instFluxAtZeroMagnitude: 1e+09). Since we are so close to bitwise parity up to this point, it would be really nice to get all the way there (especially considering the desire to feed exactly the same inputs to both middleware versions of the fgcm and jointcal external calibrations). An attempt to match up the selection of sources used in the SFP photometric and astrometric solutions between the two middleware versions will be attempted here.