Flavio, I agree with your observations. On noise reduction I do not have much experience. For my 4.2"/px undersampled subs, I find I need to be careful with noise reduction. I use an aggressive lightness mask to process only the darker areas. I use MMT on my linear integration with noise reduction applied only to the first layer, and only a minimal amount. I double check the smallest stars to make sure their brightness is reduced by only a small amount. Also I use Script>NoiseEvaluation on a preview containing a dark area. I want the post noise reduction measurement to be at most 40% less than the before measurement. My rational is that if my subs were twice the exposure length signal to noise would have been at most 40% more roughly, and I decided, rather arbitrarily, that I don't want my noise reduction to do any more than that. But all of this is an experiment of course, and I am learning what I like and what I don't.