I fail to see how their prediction was "almost exactly correct."

The prediction expected a bell curve peaking at the end of December, at around 400 7d incidence, assuming ~65% vaccinated.

We are at then end of november, ~68% vaccinated, and the 7d incidence is already above 400, and still rising.

We can only hope that they got the shape right, the timing slighly off and the peek not too off.

Where are our goalposts?

I personally think that for such a chaotic, human system, a prediction from six-months previous is pretty good if it's within a few weeks and probably less than an order of magnitude.

Probably depends on the input parameters. Lots of calls for freedom days and such as well as a general removal of measures probably changed the model conditions considerably.