[Solved] masterflat correcting to much?

Hello all,

not really a APP problem - it's doing it's job very well, but I ran into a major problem I can't get solved. Lately I'm having problems with the calibration process with my flats. I didn't change anything in my setup, the process of calibration or anything else for that matter. It look like my flats are over correcting the dust bunnies in my image train. I have 2 kinds of "donuts": small ones that are close te the sensor of my CCD, a bit dark and are visible in the master flat. The second ones are larger, have a bit less contrast, are bigger in size and are almost not visible in the master flat.

After calibrating the lights with a master dark and master flat, registering and stacking I got a really nice M101 with only 5 hours of data (luminance) with my SBIG STF8300M and 10" F5 Newton. A bit of gradient, but that's normal considering the light pollution I'm stuck with. After applying the "light pollution correction" in APP the result has strange gradients in the background I can't get rid of. After a lot of try-and-error with different settings, a new set of flats etc I finally figured out that my flats are causing this problem.

Not the way to go (I know) I tried applying the light pollution correction to my master flat to see what's going on with the flats. As you can see in the attached screenshot it's full of small donuts (dust close to the sensor) and larger donuts caused by dust on the luminance filter. The smaller donuts are corrected well, but it looks like the larger ones are a kind of over corrected and now appear as bright cirkels in my stack. To make the whole picture complete I added 4 pictures: the final stack before correction, the stack after correction with "light pollution correction", the master flat and the master flat after "light pollution correction". Thus last one is really showing the donuts. Besides APP I also tried calibration in 2 other programs with the same result. Anyone having any ideas?