So..
Griff - I have two books with two contradictory answers, so I figured that there was at least some rumour involved here ;-)
Rotorfossil says that the boundary stays roughly put. Flyting and Rotorwashed (from Wikipedia) that the boundary between driven and driving regions will move outwards as the collective is lowered, thus increasing the size of the driving area. Which makes some sense, as collective lowered->more RPM.
But I don't think this second view is necessarily correct. The angle of attack reduces along the blade from root to tip, and it is this reduction that produces the stalled/driving/driven regions. The point of equilibrium occurs at a particular angle of attack so, if you lower the collective, this point will move inwards along the blade. If that's not clear, consider the original point of equilibrium - its angle of attack is reduced once the collective is lowered, and this will push it into the driven region.
Given this, the increase in RPM associated with lowering the collective must be either the result of the reduction in size of the stalled region (which I'd not considered - thanks, RF) and/or an overall reduction in drag from the reduction in pitch along the blade.
What have I missed?!
--Dave