This has been much discussed. I think Tim's (Baker) latest theory is that extra limb mass increases "hysteresis," a term used to describe essentially what Steve calls virtual mass, although it is measured in efficiency terms versus arrow mass terms. Basicially the idea is that extra limb mass causes the limbs to bend or vibrate upon the loose, robbing energy to friction (the internal friction of anything that bends).
The most interesting realization I made during the discussions was this, the kinetic energy related to (extra) limb mass IS transferred to the arrow at the end of the power stroke. The idea being, the limbs kinetic energy is delivered to the arrow at an increasing rate as the angle the string makes with the arrow is increased, this provides the mechanical advantage necessary to "leverage" the relatively high mass of the limbs at low speed to the relatively low mass of the arrow at high speed. Essentially, the string angle iacts as a set of "gears." As the limb ends speed approach zero, the arrow nock speed is still accelerated due to this gearing. This tends to muddy the waters when it comes to judging which is the best, low mass or high energy storage.
But when it's all said and done, lower limb mass is "better" in all practical applications. The old saying that a longbow throws heavy arrows "better" (than a recurve) is a misconception. Yes, it throws heavy arrows better than it throws light arrows, just like every other bow. But it doesn't throw heavy arrows better than a more efficient bow or higher energy storing bow (ie. recurve).
