For the vast majority of situations I think the potential ability method is the way to go. There is a small number of situations where it isn't, but I'm not sure it warrants the extra programming that would be required.
I'm actually going to take an example from another sport, cycling. Lance Armstrong was a world class cyclist (including winning a world championship in his early 20s), but because of his build he was never going to win the Tour de France. Of course, after recovering from cancer his build and physiology had changed to such an extent that he was a completely different kind of cyclist. The result is that he was able to go onto win 7 Tours (de France). One might argue that his potential was increased by the cancer, but it wasn't. The potential was always there, but required major changes to be reached. Without the cancer he would have been successful, but probably no-where nearly as successful as he was following the cancer.
I think it could work if major events altered potential slightly, but probably only negatively or selectively on certain characteristics. So, a player who suffered a couple of severe hamstring injuries may have a similar or slightly reduced potential, but his injury proneness would increase (due to torn hamstrings being problematic in the future). Also, the potential gains in his pace and acceleration might decrease. The injury shouldn't necassarily affect his passing or anticipation or whatever other potential, but could certainly have an impact on certain areas.