Hello,
Obviously the right answers is "it depends".
More usefully, in general I have less than 1% of rejection. But this is not the key criteria. I mostly look at the "rejection-high" (after STF extending it). In general it has a random pattern of pixels (a few hundreds, say) - if it is all black or looks like strong noise you are eliminating too few or too many pixels.
In my area I cannot take more than a few pictures without having a trail of a plane or satellite. They can be clearly seen on the "rejection-high", even if not obious on the images. In these case I take a preview on the "rejection-high" to cover the trace (or part of it), carry it on the resulting image and examine it magnified. You want the elimination to be high enough to hide the trace, but not much higher (you loose S/N ratio when you eliminate too many pixels). In two or three iterations I get the desired result. You can monitor the calculated S/N ratio to check that you are reasonable.
Between rhe RGB I have very similar sigma high (typically 2 to 3), I use the same one unless one has a difficult plane trace. The L has a somewhat smaller sigma, usually because there is more of them. I look at the percentage only to see if it is reasonable (if very near 0 or much higher than 1% I have a second look).
For the low side I do not know, I leave it a a signma of 3 or 4 and almost never have pixels eliminated. I am not sure this is right.
Hope this helps.
-- bitli
