All those incremental changes is what made my research work indeed. As we described in the paper, the margin we had on amount of signal (dependent also on the conversion efficiency!) was small, so every % of loss anywhere in this chain of photon from emission to detection mattered.