Okay, I have an explanation, but far from an understanding. The problem is that I checked Clip Low Range in the first pixel rejection tab, and set Range Low to zero in the second pixel rejection tab. IIUC, this setting means that mis-registered areas of individual subframes - ie, the black areas caused by dithering, slight rotation or centering inaccuracies, etc. - will be excluded from the stack. Since only frames that actually have data in those areas will be included, the stacked result won't have those near-black areas or jagged registration edges. The data in the non-overlap areas won't be as high quality, because there isn't as much of it, but at least something will show there rather than zones that are basically black.
But I have no idea why this changes the background level (ie, the place where there actually is data from each frame, but it's only background sky) so much. If all the subs show 0.02 for the background at the same spot, why doesn't the final result show that as well? I also don't understand why a single frame (number 23), aligned just slightly different from the others, is the trigger here. Each frame has its own alignment, and there's nothing particularly unusual about this one.
Any ideas?
Kevin
