I would say that AM is a good measure of players' performance because it is relatively simple and it tells you a lot about what happened on a tennis court during a particular match, however this is not an absolute (objective) measure of the match quality (well, perhaps it is, but not a perfect one).
Let's assume that two players played a tennis match. After the match, they calculated their AM's (30% and 20% respectively).
On the second day the players played the second match. The first player performed better than on the first day, so did the second competitor. After the match they calculated their AM's which happened to be exactly the same as in the first match.
Finally, on the third day both players performed better than ever, however their AM's were still the same as previously.
Overall AM's are not always a sensitive measure of player's performance.
On the other hand one could argue that this examples are purely theoretical, without any significant relevance to real tennis matches.