I missed part of 2 and all of 3 for the same reason. I thought you had to "get" to the scenarios to actually consider them.

So for instance, I thought all choices to 'not play' were equilibria since all other choices are moot. But one of them wasn't even though it's not essentially part of the tree any more.

And for the other, I thought you assumed you were getting to the subgame in the first place in order to then find its equilibrium.

So the lesson is, even stuff that isn't going to possibly occur still needs to be considered for analysis.