I don't think it's necessarily a bad idea as long as their is significant written justification for the change in review score and it's not just a stealth edit. I think it's preferrable to treating reviews like gospel and reviewers like saints.
They can make mistakes and should be allowed to change the review for the benefit of people who may not have purchased the game yet. For example: there is no doubt that plenty of games these days get gameplay and longevity marks just for having multiplayer when there's no chance for the reviewer to have invested enough time to appreciate the subtleties and the quality of the multiplayer when the review is first posted.
Mariokart is a game that I would have put in the low 80's after a weeks play but probably over 90% after 3-4 weeks play.