Even after editing, it is still wrong. Individual votes is a complete different category from straight votes. These have to be separated from the straight party votes to create the y axis. Your y axis is not the same as Dr. Shiva's y axis in any way or form.

There are two options. Straight ticket or individual vote.

Those that vote straight ticket will very unlikely change their option.

Dr. Shiva explains it well.

Imagine 200 votes in a precinct.

100 are straight party.

100 are individual (you are neglecting these from what I see).

Out of the first 100. 60% vote for Republican (Shiva says all 60% vote for Trump) and 40% vote for Democrat (all 40% vote for Biden).

Out of the second 100 (individual), 65% vote Trump. 35% vote Biden.

The point on the axis is (60%, 5%).

It is only those second 100 votes that have to be considered on the y axis not the total president percentages. They are different categories.

You could try to say that, out of 60% of the first 100 votes that voted Republican, a huge percentage will not vote for Trump (I very much doubt this) so that

the axis point shifts perhaps 15% to (60%,-10%).

And this happens super consistently at a percentage that causes the curve to come down super smoothly.

This is extremely unlikely. That a large percentage would vote straight Republican and go through the effort of changing their vote to Biden when you can just vote individual. There is another possibility for the slope to happen just by having individual votes not be correlated to the straight party. Still does not explain the whole ordeal.

Once again, the y axis are extra vote % only from individual voters, not an overall trump vs. democrat ballot like you are doing here.