UZR is accurate if you use it properly. It tells you how many runs a player has saved in a season, compared to the average defender at their position.
Like any counting stat, it's not necessarily predictive (representative of the player's true talent). We don't complain that wRC isn't predictive, why complain that UZR isn't. Colby Rasmus posted an 11.2 UZR last year. He saved the team over a win on defense, compared to the average centrefielder. But there's no guarantee he'll continue to be exceptional of defense, just as him posting a 130 wRC+ last year doesn't solely guarantee he'll continue to rake on offense.