It shouldn’t have any overshoots. What it’s doing is using progressively "cheaper" non linear filtering, and some other optimizations in the audio path. Until you get into the low CPU modes which are using a less CPU intense clipper (but based on the same technology) that isn’t quite as clean as the full modes, basically because it’s doing the same amount of clipping but in a "cheaper" way as well – done for the purpose of reducing CPU use, not latency so much.
the disadvantage is a bit lower quality sound.
understand? or more confused?