Good question on goal timing - the standard narrative on (say) Gerry Cheevers was that he was better than the numbers showed because he didn't give a damn about late goals in blowouts. Caveat: I'm offering the theory as stated, and have not tried to verify it.
I think that this is valuable to track, and orthogonal to the first measure that you're tracking (so tracking both would be good, but keeping them separate would be good too). It's related to the concept that Bill James offered in the early 1980s called "victory-important RBI", where he attempted to distinguish between (putting it into hockey terms) a game-tying goal and a goal late in a 8-1 game.
This could be potentially done objectively with data, for which I'll propose a method - for each score differential and time situation, calculate the change in win (points) probability that results from a new goal being scored. For instance (and making up the numbers) if the home team has a points expectation of 1.1 at the start of a game, and they score a goal two minutes (making the score 1-0) raising their points expectation to 1.4, then that goal improved their points expectation by (1.4/1.1-1) = 27%.
Once you have that structure in place (and it probably exists out on the internet - I've built some on my own but not for serious use), then you could call "important goals against" as those that changed the goalie's team's points expectation by more than a given amount (say 20%). Goals in close situations late would be magnified, and goals in blowouts would not matter.
Anyhow, I've described an objective approach but a subjective approach could be useable too.
