Two separate issues here. One is Monte Carlo. 538 is a simulation, so the "forecast" is made by running the simulation a few thousand times and totalling up how often Obama wins. In that sense, the forecast takes into account the possibility that the news cycle will go in unpredictable ways and things might chance in ways we don't know.
Second is margin of error. 538 says that Obama's chance to win the whole thing today is 70%. Princeton Election Consortium says his chance is more like 90%. However, they project Obama winning the same number of electoral votes -- 290! How is it that they agree on this point and disagree on the chance to win? The deal is that "chance to win the election" is really better phrased as "chance that Obama's number of electoral votes is at least 270," so your level of confidence on that is directly tied to how exactly you think you can predict the electoral vote outcome. PEC believes that their method has less error than 538, so the horse-race odds end up looking very different -- but they're actually showing basically the same state of the race. (And, nota bene, it's a state of the race where the median result is a solid Obama victory). Another way to say this is that they have the same data, but Sam Wang thinks Nate Silver is hedging unnecessarily in his conclusion from that data, and Nate Silver thinks Sam Wang is being overconfident in his.
Personally, I think Wang's perspective has several things to recommend it, but Silver's the more conservative conclusion and hence less likely to be wrong or argued about. I'm probably more confident than 70%. Maybe not quite as confident as 90%, but it's tough to be sure.