That graph of Nick's above is quite clever, and you can use it to prove this point. If you hold a pencil against a fixed radious of turn and then tilt it to simulate different groundspeeds, you can see the angle of bank changing in the turn, and the angle of bank is indeed greatest when the groundspeed is highest.
A few people have dabbled on this thread with trying to look at secondary aerodynamic issues when rolling into a turn. I'm quite interested in that, although I suspect the answers are complex and maybe type specific also. I have a firm belief that the R44 pitches up on rolling into the turn, and then pitches down as the turn develops. In all my experiments I've tried as hard as possible to move the cyclic purely laterally, but of course I can't rule out user error !
Anyone care to confirm, counter, or attempt to explain ?