Because at most physics grad schools, test scores will never be a tie breaker since they don't really care if you score beyond a certain point. That varies for the PGRE at different schools mostly based on whether or not they have quals. The PGRE has nothing to do with your potential as a physicist beyond a certain score. It does however predict if you will study and pass the qual. However, a lot of schools have started to phase out the written qual. Chicago and Stanford have completely eliminated. It was specifically on my letter of admission and confirmed by my friends who go to these schools. Even MIT now has only one required qual, the others can be substituted by courses. Part I is officially eliminated. Several of my close friends passed it right before it was phased out. However if you go to a school that is less well known and/or less rigorous, the PGRE can help gauge your preparation for grad school. However, I have seen people with lower PGREs from small schools get into top ten schools if they have done great research, especially in REUs.
In grad admissions, the tie breakers are letters and research experience as well as course rigor and performance. They will take someone with an overall 3.6 (if there is a strong upward trend) if they have outstanding letters and research. I can confirm this with several examples.
The GRE a formality that basically comes from the overall graduate divsion. Your application usually goes through them. For example, at Harvard, Stanford, Michigan, your application goes through the graduate school of arts and science. That is where the application comes from. Like I said, in physics it only matters if you score so low that it suggests you don't have the mathematical preparation and/or basic reading comprehension, and/or can't write a coherent sentence. At a presentation one of the speakers pulled up a plot of GRE scores highlighting the ones who were admitted/rejected. He asked us if we saw a correlation and then promptly said that you don't because there isn't any.