Can someone cite a source for the idea "exercising" reduces induced set?
I get the idea it's one way to avoid missing desired weight. But in my experience exercising a bow prematurely (which if you aren't finished with the tiller is, by defininition, prematurely) actually leads to MORE set. My thinking, to reconcile my experience, is that when the tiller is not yet perfect you aren't actually working all of the limb(s). Rather only the parts that are bending are getting worked, in fact worked harder than the surrounding, more stiff wood.
I try to draw a bow as few times a possible during the trial/error stage of tillering. Then once I get the limbs working together, sharing all the load, I sweat the bow and fine tune shape and tiller to final weight and draw lenght. I usually don't loose any weight to 20", then loose between 5# and 10# depending upon the project as I work it up to 28" or whatever. Similarly, I don't see any set until 20" of draw, and then see 1/2" to 1 1/2" of set in the last stage of tiller and workout.