The claim is that you would always find the negative slope Dr. Shiva highlights (to show that this is very likely the result of a programmed algorithm) due to mathematical constraints. While the constraints are true at certain levels, the assertion that these constraints necessarily lead to data always showing a negative slope is wrong. As an example, take the hypothetical case where individual votes are made in the same proportion as straight party ticket votes in the different districts in the county at different percentages (i.e. some districts lean more republican or more democrat by varying amounts and individual votes are cast in those same proportions). In this case you would see a straight horizontal line crossing the y-axis (Trump - Straight Republican Party) at zero.

Edit: adjusted the hypothetical case thanks to a comment from a more informed Pede. The result is the same in that there are scenarios that can produce a straight horizontal line at zero. Edit: apologize for potentially calling pedes shills undeservedly. We should have a discussion to get to the truth.