GCP quality/selection question (aerial survey) - also optimization

[andyroo]

Question: Is it better to ignore the GCP selection for blurry or faint GCPs and only select a few clear ones (maybe 3-5 out of 10-20 images) or to manually position all GCPs and "guess" at the blurry ones (or all but the few blurriest)?

If I knew more about how GCPs are averaged, and how whether or not they are selected affects alignment optimization and dense cloud generation (or re-running the alignment step) then this might help me answer myself...

What I do now:

For GCP I first run align then manually position GCPs after import. First I position 3 - 5 points throughout the flight area on every photo where they show (or all but the very blurriest few), then I update reference, then select additional GCPs, occasionally pressing [update]. At the end (and after removing error points with gradual selection tool and manually), I [optimize] alignment before dense cloud/mesh generation.

What I wonder is: Would my error (or noise) decrease if I do not select the center of blurry GCPs?

Also a related question: When selecting GCPs, does it degrade my alignment if I [optimize] several times? Generally I try to keep my optimization step down to once or twice because I am unclear how the optimization affects the quality of initial alignment.

Thanks for any thoughts, and Happy New Year for those on the Gregorian calendar!