Your nationwide numbers are also off for several reasons. For example many people voting in 2016 skipped the 2020 elections for various reasons like death. Further pooling including exit polls doesn’t match up with actual voting very well suggesting a significant portion of Trump voters are unwilling to say they voted or plan to vote for Trump. That or a truly massive pro Republican voter fraud which seems unlikely.
Anyway looking at hard data. In 2016 Clinton got 65,853,514 votes Trump got 62,984,828 votes. A Democratic margin of 2,868,686 million votes. In 2020 Biden got 81,282,896 votes vs Trumps 74,222,484 a Democratic margin of 7,060,412.
Using the middle of your range 62,984,828 * 0.075 - 65,853,514 * 0.035 that’s a net gain of 2,418,989 votes which far less than the actual net gain of (7,060,412 - 2,868,686) = 4,191,726 votes that Biden over Trump.
To gain the (4,191,726 - 2,418,989) = 1,772,737 votes from a 3 million 3rd party voter drop would have taken a truly extreme margin of ~80% even assuming your other assumptions where correct. That might be reasonable if they where mostly green voters, but it’s largely from the Libertarian party.
Biden increasing his vote share at the expense of third parties doesn’t indicate any structural benefit to Democrats from higher turnout. It seems quite specific to Trump/Biden as candidates versus Trump/Clinton. Clinton and Trump had very high disapproval ratings, which pushed people to a surprisingly strong third party vote. Biden had much lower disapproval ratings than Clinton, which led those third party voters to vote for him in 2020. That doesn’t show a structure benefit to Democrats from higher turnout.
I’ve done a little numeric modeling, and it produces some surprising implications about who new 2020 voters selected. Pre-election polling showed that 9% of Trump 2016 voters planned to vote for Biden, while only half as many Clinton 2016 voters planned to vote for Trump. Moreover, they showed people who voted third party in 2016 strongly favoring Biden.
Trump won about 11 million more votes than in 2016, and Biden won about 15 million more votes than Clinton. But there was an over 3 million vote drop in the third party vote. If you assume that Biden pulled in a significant number of Trump 2016 votes, say 6-8%, while losing only 3-4% of Clinton 2016 voters, and won the majority of people who voted third party in 2016 but not in 2020, then its quite likely Trump actually won a majority of the 22 million new votes in 2020.