Dietmar,
Well, I’ll give a best I can answer to this.
What I don’t want however is to start a StarTools vs other software strong debate, or to in any way
Insult or abuse, or offend StarTools. I bought StarTools as soon as I saw it available in 2013, and have used it in some part in literally every image I have processed since. And also, I know absolutely from a scientific aspect, I am not retaining integrity of data in the way the author of Startools intended by the way that I personally process my images. i accept this fully. It’s just, in my eyes, aesthetically, I prefer the finished result I achieve by the composite method I use (which does change from image to image depending on subject and quality of data), and also I already have to use other software to stack the images into a final integration, and already owned those other pieces of software.
I also use a starless processing method on nebula later in processing, and find this much easier to do in Photoshop using plugins.
So firstly in my explanation, comes the usual initial processes on an image; gradient removal, colour calibration, and the initial stretch from linear data.
I own and use Astro Pixel Processor for calibration and stacking. It can save out in 32bit floating point FITS linear, straight into StarTools if desired, but APP has a very effective gradient removal, and star colour calibration tool built in that I can choose to use (and do) instead of the Startools Wipe and then later colour modules (I however, do often use colour module in STools later sometimes as well.)
There is also in APP a very good initial stretch, which personally I find gives a smoother , better tonal, a seemingly larger dynamic range, result than either of the StarTools stretch modules. May be subjective, and I KNOW this compromises the data integrity arguments of StarTools use, but there it is, I just like the stretch result better from APP.
So then I load into Startools as a nonlinear SRGB file, and do a Bin, usually 60% or 70% bin (for my imaging setup), followed by contrast, HDR, Sharp, Svdecon ( which cannot be matched in results by any other software).
Then sometimes colour module if tweaks to colour balance are needed (basically balancing the histogram), then entropy, and finally super structure (usually saturation, maybe brightness, and a little dimsmall- 20-40%)
Then I save out and pull into photoshop and remove the stars using StarXterminator plugin, and follow with a hefty noise reduction routine using NoiseXterminator ( I know its AI, and I have not got Startools tracking any more), and after maybe a few tweaks using ‘camera raw filter’ replace the extracted stars with ones extracted from only a lightly stretched and star colour calibrated but otherwise un processed starfield from the initial APP save out. i do this because I just prefer the star look using this method.
Its then usually saved out as tiff and pulled back into StarTools for a light flux effect and thats it.
For publishing I then bin again (a copy) in Startools to 2048x ???? Which I find is a good image size for ipads, web use etc.
So, in summary I use other software with StarTools because
a) I prefer the APP initial stretch.
B) I TOTALLY accept that my scientific data has been compromised in the pursuit of an artistic image.
C) Overall I prefer the stars extracted from a lightly stretched and well saturated image plugged into a starless well processed and noise reduced object image, so I ignore what stars look like in Startools, push the object processing hard, then remove and replace the stars in Photoshop.
This, after processing using this type of method gives to my eye better results than purely using StarTools modules alone.
That is why I don’t post much, I feel a little guilty, and I don’t want to offend StarTools purists, believe me I LOVE STARTOOLS, and i use it in ALL my images! (Just not as the brilliant Startools author intends).
I add some recent pics, but had trouble with posting so low quality and not HDR looking.
Andy.
maybe later ... 
) to identify stars. Why not? ST's star mask does just the same, but by using a direct mathematical approach. (BTW: I even don't like the term AI - for me all this models are still just neuronal networks. But that's another story.)
And NGC 4217 also looks great. Impressive for not even four hours. I once tried M 106 and aquired more than five hours but struggled. Bright moon might have had something to do with that.