I've always felt whip ending should give more speed, despite plenty of opinion to the contrary.
If you take a stiff limb tip and say it takes say 10milliseconds to recover from full draw, if you then remove material from the the last say 5", and draw it to the point where that point 5" down the limb is at it's previous deflection, you will actually be drawing further (due to the additional tip flex).
The point 5" in should still take 10 milliseconds to recover (as you havn't changed anything on the handle side of that point), but while it has been recovering the last 5" has recovered the extra draw (assuming a 5" limb with very small deflection will recover at least as quick as the remaining 20+" of limb).
Thus your longer draw has recovered in the same time and is therefore faster...now factor in that you have removed mass, even quicker still.
Off course this my be an over simplification and total bull, because in reality you would have lowered the draw weight of your existing bow at the same draw...but if you do it before the bow is reduced to final weight to get the effect at your required draw length the logic should hold good.
Presumably the reason a Holmgaard design works is that the stiff levers are light as well as stiff due to their narrow/deep section.
Common sense would tend to say that all parts of the limb should work together for maximum efficiency (ignoring the string angle/recurve arguments)
Del
(blimey, I need to lie down now)