You're comparing apples and oranges in many different respects.
a. Trump Vote =/= Biden Disapproval. Not everyone who voted for Donald Trump did so because they hated Joe Biden, some just supported Trump more. Initial approval ratings of presidents in their first 100 day grace period are usually higher than their share of the vote (Trump being an outlier), likely because people who didn't vote for them are keeping an open mind. Obama's initial approval rating among Republicans was 43%; I guarantee you half of Republicans did not vote for him.
b. Voters =/= General Public. Only about two-thirds of eligible voters participated in the 2020 election. The electorate is in general older, more educated, and wealthier than the general population. It wouldn't be surprising that Biden would poll better among all Americans than voters specifically.
c. Improvement =/= Winning. Donald Trump definitely improved on his share of the minority vote, but he still only received 26% of the non-white vote. He improved from very, very, very unpopular to very, very unpopular. That was newsworthy because even small margins like that can swing states with high Hispanic and Black populations, but it's not as far off from Biden's disapproval numbers as you might think.
All of which is to say there's no solid evidence of an actual shift. That said, there was a very real documented polling collapse for Donald Trump after the January 6th attack in a span of merely days. It wouldn't be that much of a stretch to assume that could have translated to higher Biden approval as well. But the data just isn't as clear-cut as you think.