I don't think throwing out Utah data is appropriate because it is reality, but the suggestion you have to treat them as 1 data point (i.e. BYU went 0-1 against Utah over that time) could work. It would also require doing the same for a few other teams (see below*), and realistically isn't my preferred way to handle the data, but is pretty easy to do. It would have the unfortunate result of not allowing for comparisons as to how BYU does against different levels of P5 teams though (at least not without more updates than I want to do, since Utah and other schools BYU has played multiple times sometimes fall on the good side, and sometimes on the bad side of that ledger)

*Additional updates would be that
BYU went .66-.33 against Texas (instead of 2-1)
BYU went 1-0 against GaTech (instead of 2-0)
BYU went .33-.66 against Wisconsin (instead of 1-2)
BYU went 1-0 against Arizona (instead of 2-0)
BYU went 0-1 against Notre Dame (instead of 0-2)
BYU went 0-1 against UCLA (instead of 0-2)
BYU went 0-1 against Washington (instead of 0-3)
and, of course, BYU went 0-1 against Utah (instead of 0-8)

Overall that would change BYU's P5 record from 16/42 to 13/24, which would have them profile as more of a middle-of-the road P5, but with a much more uncomfortable sample size. Here we can also no longer easily discriminate between performance against good teams and bad teams.

This also minimizes the impact of 3 of our 5 best wins of the last decade (2012 GaTech, 2013 Texas, 2018 Wisconsin). By this metric instead of having 5 quality wins against good P5 teams, we'd be down to 2.66 over the last decade (our 2012 GaTech win only counting as half a win and our 2013 Texas and 2018 Wisconsin wins only counting as a third of a win for each, in exchange for each Utah loss only counting as an eighth)