I had this similar problem and now it seems to be better with QPC turned on and "preserve PDC delayed monitoring" on. It was probably 20-30 ms early before. Anyway it was enough that quantize to 16th notes always messed things up if I didn't move them forward by hand first. Now it seems to work fine with quantizing. I hope I'm not dreaming this
BTW: my ASIO latency is usually 64 samples (1.4 ms)
I have to do more testing, I hope I won't get the flu that is running around in my family... I'll get back to this if it turns out that I was imaging things.