Depends on how you do the poll. If, as most pollsters do, you say at the outset, “We will interview 30% Rs 35% Ds, and 35% Is,” then you would be correct.
Another way to do it is to randomly get 400 likely voters and go with what the numbers tell you. Now, you MIGHT have gotten unlucky and by accident have gotten “too many” GOP, or in fact the randomness may have picked up a shift.
You may be right, but changes that end up with an eight-point shift in D/R/I breakdown is not typical by any means.