Recall that in 2016, Hillary Clinton won the national popular vote only because of her margin in California; if you subtract California, Trump actually won the national popular vote (as well as the Electoral College).
In any national poll, the pollster will include a proportionate number of respondents from heavily populated -- and heavily Democratic -- states like California, New York and Illinois, thus producing the Democratic nominee's lead nationally.
Right now, the RCP national poll average has Biden at 50.6 and Trump at 42.5. (This includes the WSJ poll reported today, which has the smallest sample of any poll conducted in the last two months and includes only registered voters, not likely voters, and does not include the Zogby poll reported above.)
The RCP polling averages also show Biden ahead in all but one battleground state (Texas). And yet, the same RCP polling averages shows Trump running head of his 2016 polling averages in the battleground states (by 0.6%).
So, where are Trump's numbers coming from? If he's over-performing his 2016 battleground numbers -- an election he won -- but headed for defeat now, you'd have to believe 1) polling accuracy has improved substantially in the last four years or 2) Trump is running up massive leads (say, 80 to 90 percent of the vote) in places like Arkansas, Utah and South Dakota. After all, that 42.5% lives somewhere, right?
Either of those things is of course possible. But they both seem ... unlikely.