Actually I have a very good understanding. If looking at YPC your population is carries. Tossing out a full game of carries is not tossing out an outlier, it is tossing out a large chunk of the population, good, bad and in between. That would be a terrible statistical error. If you were looking at yards per game and you had one game where he rushed for 2 yds, and you found out it was due to a torn ACL, then you would right to throw out that game. But I cannot see what statistical sense it would make to throw out his best game of production even if your population was yds per game. What would the anomaly be to make that data a possible error?
The reason to toss an outlier is to avoid bad measurement or some sort of data error. Were any of Mills runs measured wrong? Was the data miscalculated? Did the defense lay down which produced greater production? Nope each run is a pure accurate data point of his production and would not be an outlier.
Now however I do agree that production data can be misleading. Ex. the yr Neb destroyed Fresno St at Fresno. It was big play after big play that scored points. But no consistent production. While the yardage and the score board looked awesome, watching the game, it was obvious there were issues. Mills eye test did not give the happy feel of his calculated production.
So I know stats very very well. But you can tell English and writing not so much