Although I agree with the above description of the problem, I disagree with the technique since it is implied that all of the luminance frames were aligned and sum-combined and then a dark subtraction was applied. This is a VERY serious mistake since the aligning destroys any correlation there was originally with the dark frames.
The proper sequence of events is to dark-subtract each of the luminance frames and then proceed with alignment and sum-combine. In this case, the problem would have disappeared immediately and it would not matter if there was a shift/drift in the FOV during the sequence or not.
Anthony,
All frames were individually dark subtracted BEFORE combining. As Anthony rightly says, to do anything else would be a serious mistake, but that is not what is happening. It seems that with some DSLR's (I've seen this in mine too) that for some reason that not all hot pixels are repeatable across the light and dark frames. That is the only reason I can think of why hot pixels would appear like this even after individual dark subtraction.
If you can shed any light on why this would happen I'd certainly love to know. I have examined individual dark frames and visually confirmed that hot pixels which are on the light frames are just not there. I know cosmic ray hits can cause this but not in the same place in so many light frames.
The only other reasons I could think of was that the CCD behaves differently when it is exposed to light (while taking light frames) rather than in darkness (while dark frames are gathered).
  