Putting aside reciprocity failure, I had thought that an additive head with narrow band RGB filters would be better than a subtractive head, as it would more closely mimic the RGB characteristics of a digital laser printer. I understand Drew's point re: shifts on startup and fadeout, and due to heat during exposure, but I am willing to bet these are minimal and somewhat consistent. This seems to leave the main theoretical problem as being potential crossovers due to much longer than designed-for exposure times. I am about a week away from having the time to test this, but it should be interesting.