yes, this does not answer your actual question but i think if you carefully inspect the background model extracted by DBE, it's going to be far too lumpy. check your samples and as carlos says, move them off of any stars. while the brightest part of the star is rejected by the algorithm, the 'halo' around the star can be picked up and the tool thinks it's part of a gradient. same goes for the outer arms of the galaxy. what you can do is re-place the samples on your DBE'd image - it's a lot easier to see where there is galaxy signal in that image - and then save the process icon and with the original image in the foreground, close DBE and then double-click the process icon. that will transfer the samples to the original image, and you can try again with confidence that you haven't picked part of the galaxy by accident.
i think i said this on CN but when you are shooting in LP skies, your background is composed of the unwanted signal of the skyglow. when you subtract a smooth model from that signal, you're pretty much left with the noise. you can't really see the noise until you subtract the background because it's a small percentage of the signal that's there. but remove the signal and suddenly the noise jumps right out at you.
when i shot m101 from my home, i had to put in 30 hours to get something that looked even remotely okay, and i still had to do a lot of NR on the background, and kind of clip it off.
rob