Interesting approach, Mike, I will try it in my next image. I have experimented in the past with Starsupport as weight (usually divided by a fix number of star, to some power<1), which is, I think, better than FWHM since it factors in background levels/noise.
Ignacio