True, it was. But if you dig into the crosstabs, you'll notice two things about the most recent poll:
1. It used the smallest sample size (600, compared to 700 - 1000 of the other polls)
2. It was a poll of registered voters, while the others were polls of likely voters
So it was the most recent, but also the most poorly done, with the highest margin of error. Good news for us.
It's interesting to note that "likely voters" as a category are generally older and more conservative, especially in a midterm election year. You would think a likely voter measure would produce results that were worse for prop 19, but it seems the opposite has been true. I wonder why this is...are we finally getting support from those crazy ass tea partiers?