This last comet was indeed a particularly tricky one to get right - at least in my hands- because it was moving so fast. In my F4 , f 1200 mm system I tried stacking lots of 5s frames so that this movement should blur the comet as little as possible and the centre for stacking be best defined.
However the comet-stacked images then looked like this.. i.e. stacked based on the comet's bright core
When I then used RC StarExterminator on this integrated comet-aligned stacked image - the same as in your case - star removal didn't work completely and trails were left. I think that this is because - at 5s - the star images are too close together and not resolved so they are not consistently recognised as stars by the Xterminatoe program.
My particular solution was to go up to 20 and 40s subs which produced linear arrays of more discretely separated stars rather than the almost continuous lines obtained at 5s . StarExterminator was then able to remove these better defined arrays much more cleanly.
The price paid was that the exposure time was a little longer than ideal and therefore the comet stacking wsa based upon a slightly oblate brightest centre point. But the image at the end was passable -and looked quite good after the stars were added back in (i.e extracted from the star-based stack aligned to the first image of the comet -aligned stack) . Overall this method worked and I think a bit easier +faster to carry out than running a star removal on each frame individually before stacking.
I haven't tried running RC StarExterminator on all 180 x 5s frames individually and then trying comet stacking - theoretically it could produce a sharper less movement-affected image but it would be a lot of work. I also suspect that the low SNR of necessarily short individual frames might create problems with artifacts in star removal -- and that this itself may be a problem if that is your procedure?
I realize that this doesn't really help you with processing your current data set since the solution for me was to change data acquisition exposure times but you are not the only one I am sure to have encountered a) difficulty in removing star 'remnants' from the comet-based stack of images and also b) the converse, to cleanly excise just stars without any residual spread out comet signal from the star-stack that you add back to the comet stack (this is perhaps a problem with the first image in this thread?).
Incidentally-- one easy fix for b) is of course is to simply image the stellar background just after the comet has moved on and out of the initial field.
