Earlier quoted context omitted.
Sorry you're confused. Point was pretty clear: US polling data was not found to correlate strongly with reality, so treat it with some suspicion. Can't really break it down further than that.
So saying an event has a 10% chance of happening, and then it happening after one roll of the dice, is enough reason to doubt the data?
Polling was incapable of accurately determining what the odds really were. If that die was six-side, but I told you there was a consistent 10% chance of rolling a 1, then I'm providing you with a bad probability. The media and pollsters did the same (for whatever reason).