Interesting project!
I see what you mean about the problem with range. Either a device has limited range, or... the range tolerance is ridiculous (varies a LOT with different conditions). IR is pretty much line of sight outdoors, and active both sides - although perhaps you could have a passive relfector on the hamdle bars which, under normal circumstances is close to a IR emitter/reciever (wrist). As soon as the biker moves his/her hand away from it (or tilts it for that matter, ala Erics idea)... it pings, and does what you need it to do.
You could always include a small delay, so it waits for the reflector/oject to be 'out of range' for say, more than a second, before it does anything... that way you don't end up with lights flashing on and off quickly in the rain, or if the rider twitches their wrist. (or scratches their nose/bum).
Actually depending on usage, I'm not sure the range would be a problem providing you use some form of delay/filtering.
What about a small switch on the wristband? the rider moves his wrist back to switch it. Its not a common thing to do with ones hand, and when the arm is being held out to indicate direction, the hand isn't needed for much else.
For a high tech approach, RFID tags can work at that range (inductive), even the passive ones - although they tend to require rather a lot of power (= bigger batteries = weight = big muscley bikers arm) and it seems to be over kill.
One more brain storm - sorry if it seems like I'm hijacking your idea here - eric gibbs idea, with a touch of intelligence. Make the rider flick his wrist twice. A small microcontroller checks how many times the tilt switch is pinged, checks the time between them, and decides if its a valid 'signal'. Like a crude sign language glove reader