But you say only the electoral votes matter? Why does this weird three swing states that aren't even all the same so have different populations accross the elections margin matter?

Ah, you're talking about the fewest number of regular votes that would have to change so that the electoral college would flip. in that case I'd say 42k and 77k are essentially the same number, given the large number of voters we're talking about. But yes, Biden's victory was tighter than Trumps in that calculus.

Although, in the situation you propose, 2020 would actually be a 269-269 tie. Which means any unfaithful elector could refuse to vote for their candidate and throw it to the other party. And even if no one did, then Congress would break the tie in a pair votes that would presumably go for Trump and Pence if people voted on a party line.

"Electoral votes" appears to be a term you've created to mean "the smallest number of votes it would be possible to flip to change the result of the election, if allocated perfectly", rather than what most people would assume you mean -- electoral college votes. Given your definition, you are basically correct.

Trump won 306-232 in 2016. With faithless electors, this became 304-227. He lost the popular vote. As you note, 77,000 votes in the three closest states could have flipped the election.

Biden won 306-232 in 2020 (the same margin, or better if you allow for the faithless electors). If he lost 42,000 votes it'd have been 269-269, which would have led to the House of Representative contingency, which might have elected either Biden or Trump (or ended in a different outcome, frankly). It'd take another 33,000 votes to give Trump an unambiguous win by flipping Nevada.

This is an interesting curiosity -- but for Aunt Maria getting the flu, the election could have been different! But we're in increasingly silly hypotheticals. Knowing what we know now, it's clear Clinton would have done more to target the states she narrowly lost. But the problem is that state votes are correlated with one another and so the number of hypotheticals you need to sustain to "flip" exactly those votes without turning out any additional votes or affecting the campaign strategies is pretty weird.

Even if you take the prototypical version of this question "Did Ralph Nader 'cost' Gore the 2000 presidential election by 'taking' at least 500 of his votes in Florida?" it's sort of a rabbit hole of absurdities. The answer is surely yes to the question, because the answer to any hypothetical is yes when the margin is that close. But beyond that not super productive.

In general I think most people would, collectively, analyze Biden's victory over Trump as somewhat more decisive than Trump's over Clinton or Bush's over Gore, though less decisive than either of Obama's or Bush 2004.