I'll take a stab and post some thoughts - take with a grain of salt...
First of all, you have to realize that a really bad eccentricity (or even a kind of bad one) will have effects across the entire image, not just the stars. So, whatever you are imaging will also be fuzzier and this will affect the final image. In general, it is best to discard bad FWHM and Eccentricity frames - you can't really correct for those image capture errors. How bad? That is up to you and how much of your data you can trash.
If you are going to improve things, it would be with the Deconvolution tool, but you need some really good SNR for it to work well and not make things worse. There are a few good tutorials on youtube. It can only do so much, though.
If all you want is better stars then you can also try replacing the stars in the image with the ones from the RC filter through a very good star mask using Pixelmath. Make sure the backgrounds are very, very close. But, this does not solve the issue of the effect of bad frames on your target.
My advice would be to trash the bad frames - this will solve many issues before they occur, but at the expense of SNR. As you move forward, you will 1) learn how to capture more consistent frames and 2) build the overhead into your workflow, knowing you need to capture 8 hrs to get 5 good hours (or whatever).
My .02
Mike
