It's a weird thing, watching just how much a windsurfing sail twists when there's a good wind blowing. Other sails (like the ones that 18 footers use) don't seem to twist nearly as much.
One thing to note is that the area of the sail near the boom normally has a lot more curvature than the battens near the top of the sail. It's also worth noting that a flat airfoil produces a lot less lift than a curved one. Also, changing the angle of attack of an airfoil (angle to the wind) changes how much lift is produced. So if you want absolute max power from a certain size of sail, you would want the same curvature all the way up the sail, and no twist at all. (8 cams, anyone?)
Flat airfoils (like the top batten) can't be sheeted in as much as curved ones (like the one near the boom) before they stall, so if you've got a flat top batten and a curved bottom batten, then for max efficiency I guess you have to sheet out the top batten a bit, which is where the twist comes in.
I've got an 80's era Bombora speed 6.2 sail that has almost no sail twist, and in a good wind is very prone to wanting to luff up into the wind. This is known as "the centre of effort moving backwards". It really has a strong twisting moment when the wind's blowing, and can be difficult to push back when it's determined to luff up!
My early 00's RX1 has a lot more sail twist, and doesn't want to luff up nearly as much. I haven't rigged the Bombora sail for a while, so don't know which one gives more push.
I think the answer is control -- with sail twist the center of effort doesn't move backwards as much in high wind. The sail also has a much wider sweet spot with regard to how much you're sheeting in, because there will always be at least one part of the sail that's at the right angle of attack. It also turns itself from a sail with a large area into a sail with small area in higher winds, meaning that you don't have to come in and re-rig every time the wind changes a few knots.
Fluid dynamics is a strange beast, many times the explanation the makes the most sense is completely wrong, and can be proven so with the right experiment... so muck around with your downhaul and see what happens