I adjust the U layer before doing the alg, because the way I see it, it saves time.
I can recognize how to turn the U layer faster than I can recognize which specific PLL case I have. So while I am turning the U layer to where it needs to go, I can finish narrowing down the case so I can go right into the PLL alg.
Whether you do before or after is not as important as whether the AUF is seamless with the algorithm. Sometimes it is seamless if you do it before sometimes it is seamless if you do it after.