Here's some more context from the Pew Research Center

It should be noted that the GOP's strategy could potentially backfire as well. The mail-in/absentee votes which are most at risk are those which arrive on or after election day - regardless of when they were actually mailed. Republicans are still overwhelmingly planning on voting in-person and on election day (Nov 3). However, we know from previous elections that external events can suppress in-person turnout... things like severe weather, very long lines or a pandemic. Now we've got the pandemic surging - in some places it's higher than it has ever been (including in the battle-ground states of Wisconsin, Iowa and Minnesota). It's plausible that this could deter enough in-person voters - who are most likely to be Republican - from standing in long lines during a pandemic in possibly miserable weather to shift a close race the other direction.
To be clear I'm appalled by these attempts to limit voter participation. I think our democracy is stronger when more people participate, and I'm passionate about removing barriers to voting (including widespread availability of mail-in voting and in-person voting). Just pointing out that this latest gambit could bite the GOP in the ass. Maybe.
So it is important when looking at these that people realize the group sizes are not the same
Also from Pew Research Democrats have always been larger than Republicans from the total pool of voters since 1995, after being basically equal in 1994. so that 50% of Republicans who say they will vote on Election Day is a smaller number than 50% of Democrats would be. Given that the total number of registered voters according to the Census data in 2020 is 152,666,000 that means there are 50 million Dems and 44 million Reps, a not insignificant difference. Given that over 75 million people have voted already, we have 50% of registered voters already done. This already puts the accuracy of the chart in question for me because I believe more people chose to get out early or vote by mail than said they would in this survey and given how bad the pandemic how gotten since this poll was taken that is not implausible or illogical. All the GOP machine has been arguing against was not voting in person, not about not going on Election Day, so I would venture to guess that some of that 50% (and likely some of the 20% Democrats) have changed they minds and have or will vote early in person. If we hit over 102 million voters before election day (which is not unlikely given the trend) we will clearly prove that 33% are NOT voting on Election Day. We should also realize that we've never achieved, nor will we likely this time, 100%, so we should assume 90% as a reasonable high turnout which means once we exceed 92 million then this survey is wrong.
Even if things stay as this poll suggests, those 50% of Republicans are a bit under 20 million assuming 90% turnout and you still have over 9 million Democrats voting giving a possible swing of just 11 million votes in an election with 137 million votes likely to be cast, or a bit under 10%. Given Biden's polling lead exceeds this I still am optimistic that the red ripple will not be enough.
If Trump wins it will require and electoral college Royal Flush quite more improbable than the 2016 one he drew and those votes will need to show up in the right states.