A bit more tinkering today on trying to reduce the leftover light. Sadly, nothing at this stage seems to help.
Thus, Dr. Sandquist advised me to back up a few steps and to one of the most annoying steps all over, but even more tediously. This step involves looking at the radial plot of intensity vs. radius from the center of a star, making sure it has a smooth distribution (should look like half a bell curve), and then finding that star on the actual image (one star out of >2000) to make sure it doesn't have any close companions or anything else that might screw things up.
This meant manually going through 187 stars, each of which took about a minute to do. Two hours later, I had a new set of model stars to try out. Plugging it into the script I generated a new image and magically, all leftover light was reduced to half of what it was before or better!
Feeling that there was little more that could be done to improve things, Dr. Sandquist advised me to move on to the next image and see how well things look for that.
Since I knew what I was doing this time and didn't have to go the 171 steps down the hall to Dr. Sandquist's office every half hour, I was able to do things much quicker this time. Again, I had to go through the tedious task of visually inspecting each star that the computer deemed a good candidate to model the shape of. This time there were "only" 151 of them and I ended up keeping 130.
At that point it was 5:30 and I headed off to dinner. Afterwards I finished up my Powerpoint presentation for the talk I'll be giving on Friday. For the most part, it covers everything I said a few days ago, but just includes specific numbers, bigger words, and more citations. I'll be sending that to my advisor in the next day or two for her approval and then I'll post it here.