Yes that's my main point: this plot does not prove the correlation.
If you look at a bigger picture e.g. (maybe here [1], altough it's not technically the same metric) there was a peak around 94. The peak and the relative low value before the peak cannot be explained by video games.
And that my second point: no one ever claimed that violence is only caused by video games. Clearly the blue curve is under the influence of (many) other factors, so it's very hard to determine whether there is a correlation without knowing these other factors.
But let me say it once again, it's a possibility that video games reduce violence. I am simply saying that if think this is the case because of this plot, you're probably biased.
[1] http://www.eji.org/files/juv%20murder%20rate%20drop_0.jpg