While I get it and I've used this myself, this also feels like a strange metric to me in a way. The tipping point state only matters because of the electoral college, so why not just look at the electoral college directly?
This metric gets at the question of how close an election was to going the other way, in the sense that a minor change in circumstances might have swung things towards the losing candidate. In both 2016 and 2020, for example, if just 1% of the winning candidate's voters had voted for the losing candidate, then we'd have a different candidate winning.
But that's not true in the case of 2004, for example, even though the electoral college tally was closer. That's because in that case, Kerry would have needed Ohio (or some other state that voted more Republican than Ohio) to flip to him in order to win, and the margin in Ohio was more than 2%. So we weren't really all that close to seeing a Kerry victory.
Nevada, Iowa, and New Mexico put it to a tie...and Colorado had fewer votes to flip despite the higher percentage.
Pretty sure the numbers combined amount to only slightly more than OH.