I assume the PDF is a Nicholas Whye production because it is full of additional fun stuff.
There’s a really great table at the end of the PDF which I’ve included here as a screenshot:
The key thing is columns M and N. M is the rank of the ultimate winner in the nomination stage and N is the rank in the final vote of the top-rated finalist in the NOMINATION stage. There are 6 examples where those ranks are different but there are 13 examples where they are the same. In other words, the nomination stage predicts the winner 68% of the time. That’s an interesting number. It’s enough that it is impressive how often the nomination stage gets it right but it is also fallible enough that you wouldn’t want to rely only on one-stage voting.