I'm realizing there might be an issue in be68024. It has to do with deposited energy scales being different for homogenious and ScFi variants, which is inconsistent with using the same number for lightYield/photonDetectionEfficiency.

The homogenious path uses corrMeanScale = 1, whereas ScFi uses corrMeanScale = 1/(sampling fraction) to correct the energy, but that does not help us. That is because the correction is applied after the distribution parameter is chosen based on yields/efficiencies and photon count is determined from it:

The fix might be to rescale yield using sampling fraction for one of the variants.

cc @akioogawa