After a few weeks break and working on some other material, I thought I'd have another go at this data.
I went back to scratch and instead of using the BatchPreprocessing script to calibrate, register, etc., I did it myself one step at a time. This gave me what I consider to be better calibrated frames for the steps that followed.
I did four versions today as I worked through various paths and attempts. Core to all attempts were the use of LinearFit before colour calibration and the PixelMath from Troy Piggins that Rick mentions in the previous post. I also used HDRMultiscaleTransform for all as well.
On the last two versions, I used deconvolution with a star mask protection to sharpen the image more. And as I did it, I was aware of Rick's warning, but chose to ignore it this time.
The final image is here:
http://astrob.in/118616/H/ The other versions can be seen there as well (version E, F, G).