This has been used by the police, at least in Germany.
The general problems with this approach are the poor resolution of the data, giving you too many false positives in populated areas, and that you have to know the time of the event - hard to know without witnesses in sparsely populated areas where the resolution would be sufficient.
The case where I know cell phone location has been used was where a woman had been killed in a car by a tree trunk being thrown from a bridge on the autobahn. Here you had a sparsely populated area (the fast-moving people in the cars are easy to rule out) and the precise time of the event (her horrified husband was the witness).
So they got hold of the suspect, a junkie who happened to be around. They got a confession from him - by withholding him from drugs until he confessed, so I'm not entirely convinced that he really was the perpetrator. Which illustrates a big problem with this approach: The police has a tendency to make people confess somehow as soon as they have a 'convenient' suspect, and cell phone tracking can give you plenty of them, guilty or not.
BTW: Only 20% of the killers in the US are caught and convicted? It's > 95% here in Germany...