If the polls they collect have flawed methodologies and flawed samples, the RCP average is useless.
Further, without any corrections taking into account how decent or bogus the core poll results are, RCP's numbers will always skew towards the results from the most egregiously flawed poll they include in their - again - very simplistic calculation.
It’s been said cut in half any poll numbers of Bidens.
Polls also only measure “Likely Voters” That means people that voted in the last two elections. There is considerable evidence that Trump is attracting a significant number of unlikely voters, people that haven’t voted in awhile. This would considerably skew the results of any poll.