Noise estimation is a good estimator of the quality of your stacking process. In our example, the red channel has almost 1.5 times more noises that green or blue. That probably means that DSLR is unmodified: most of red photon are stopped by the original filter, therefore leading to a more noisy channel. Then, in this example we note that high rejection seems to be a bit strong. Setting high rejection to [math]\displaystyle{ \sigma_{high}=4 }[/math] could produce a better image. And this is what we have in the image below.