Hi Seb (and thank you Burly and Rowland!),
Thank you for your interest in StarTools, as well as for uploading your data.
Wipe is indeed similar to DBE or ABE, but with some distinctions and subtle differences in purpose and algorithms.
The sort of gradient elimination that Wipe and ABE/DBE performs comes down to creating a model of just the gradient/light pollution/etc. and then subtracting the model from the image, so that you are left with the "correct" pixel values.
As Rowland alluded to already, there are different ways to establish that model, and there are even subtly different ways of subtracting that model.
Without knowing what DBE/ABE produces (so I can address the issue of what you expected vs what you got), I can at least try to explain what Wipe is doing to your data (and why).
First off, as opposed to ABE/DBE, Wipe is extremely sensitive to dark anomalous data
- pixels that are darker than the true interstellar background you recorded. Such pixels can be stacking artefacts (as is the case with your data), dead pixels, dust specks not caught by your flat frames, or even terrestrial foreground objects (e.g. trees, mountains). In extremely noisy data, a dark pixel can even be due to be sheer misfortune of no photons registering (due to shot noise).
The reason for this treatment is a conscious design decision; Wipe was programmed to *never ever* subtract more than the background allows; you will see Wipe back off in areas that contain the anomalies, actually leaving in enough light pollution/gradient to make sure that that the anomalous dark pixel doesn't clip (e.g. goes below 0) when the model is subtracted. The result is halos or gradients at borders (where stacking artefacts are).
The rationale is one of detail preservation. You may not care about the extremely faint detail or nebulosity at this point (you just want to get rid of the gradient), but getting the amount to subtract wrong (by misplacing a sample in DBE for example) can cause extremely faint nebulosity to be subtracted along with it (and be lost). StarTools is all about signal preservation and being smarter with your signal - because its target audience doesn't have any to spare.
That's why Wipe is more "neurotic" about artefacts - signal preservation takes precedence over convenience in this case; it is better to know find the root cause than to just pretend all is well.
Can you tame this neuroticism, even if you completely ignore the root cause? Sure. You can tweak Wipe's settings, tell it to disregard areas, etc. You should be able to get something half decent out of Wipe no problem.
However, it is important to understand Wipe wasn't designed to work around problems of your own making - Wipe was designed to get the absolute most out of the cleanest data you (and anyone!) can reasonably muster. Specifically, this means, StarTools expects data that was calibrated with flat frames and dithered between sub-frames. They are basic acquisition "musts" regardless of what software you decided to stack and process your data in.
The gradients in your data look somewhat "suspicious"; they are very local and undulate with high frequency (something else that Wipe is very sensitive too; high frequency gradients are typically nebulosity, not unwanted light!). Sometimes this signature can be caused by condensation ("foggin up") on a part of the optical train.
I hope the above helps, and do let me know if you have a particular end-result you like in PixInsight that you would like to achieve in StarTools.
Below, a conservative (no HDR, sharpening, etc.) "textbook M42" created with your data. The colouring (default settings using the Color module) is particularly interesting in M42; it's teal OIII-rich core, red Ha emission, blue reflection and Hb.