Thanks all.
I will make a prediction that the in camera noise reduction will beat any amount of bias and darks no matter how carefully taken. I think Canon apply the noise reduction at the Bayer level ie pixel for pixel in the cmos chip. Not the derived pixels which we see as some calculated value from a pixel and the pixels around it to simulate a colour. I am just going on the results I am getting for frames with in camera noise reduction on and just correcting for flats.
This is a crop from the original Tiff (70+MB) from the Lagoon+Trifid with the 100ED from the picture in this thread. Its had five cycles of adaptive Richardson-Lucy.
http://members.optusnet.com.au/~trle..._dd_m_crop.jpg
And here is a crop of a jpg (high quality) from a 73MB tiff of a single frame. I have circled the some of the residual noise with the in camera noise reduction on.
http://members.optusnet.com.au/~trle..._1511_crop.jpg
As you can see it is minimal and when stacking disappears if images are dithered or drift slightly due to even slight error in polar alignment and output of final stack as a median rather than average in RegiStar.
Bert