You are right visual inspection is very open to different people opinion of correct shape but this is random uncertainty which is common in all testing and inspection.
The main problem I can see is the range of media as you identified can potential be massive, my boss found similar results to you Socrates.
I believe that the range is far to big to have a numerical figure depicting the total number of unacceptable particles. It should be a percentage that is based upon the screened sample currently being inspected.
