CA, for example, defacto legalized when they voted for medical mj in 1996. Violent crime has fallen by half in CA since then.
http://www.disastercenter.com/crime/cacrime.htm
http://www.disastercenter.com/crime/uscrime.htm
http://en.wikipedia.org/wiki/List_of_motor_vehicle_deaths_in_U.S._by_year
So... we should be keeping people with criminal minds stoned all the time, because pot has the well-known effect of inhibiting initiative?
What I see in the graph is not a correlation between pot legalization and crime, but a clear relationship between the lax attitudes on crime that began in the 1960s and the more strict measures that started being implemented in the 1990s when people were concerned over high crime rates.
You have to be careful when looking for a causative relationship between correlated observations. Unless a causative mechanism can be established, you might be looking at things that aren’t even connected to each other.