Imagine that you've got a cam lobe with the usual "sine-wave wrapped around an axle" kind of shape. (it's not that, actually, but the image is helpful)
Now instead of that, imagine a dentil molding, like you'd find in expensive crown molding or something, that looks like square teeth, wrapped around an axle.
That's the difference between a "mild" lobe, and an "aggressive" one.
OK now imagine the fraction of the time, that on a typical "street"-ish cam with the former style of lobe, that the valve will be spending anywhere near full lift. Pretty small, eh? Most of the time it will be lifting off the seat or gently dropping back down.
Now imagine how, if you have 2 cams that are IDENTICAL IN EVERY WAY except that one has max lift of say .500" and one has max lift of say .750", how LITTLE of the time the actual lift that's in effect, will be above whatever that "maxed out" point is. MOST OF THE TIME, the valve is going to be .100", .200", .300", etc. off the seat.
OK, so it's pretty easy to see that the cam that has .750" of peak lift will have .150" when the other has .100", .300" when the other has .200", .450" when the other has .300", and so on. Eh?
Does it REALLY matter that the "peak" lift doesn't flow very much more, when ALL THAT TIME ON THE FLANKS OF THE LOBE will have so much more flow? In fact, if you took that .750" cam and "squared off" the tip of the lobes, such that it got up to say .600" just as quick as it did before but didn't go all the way to .750", it would flow pretty close to the same as it would at .750"; right?
Problem is, cam lobes don't work that way. They are designed using a series of curves, where there's one curve on the ramp, another as the rate of rise gets steepest, another over the top, and so on; might even be different on the rising and the falling sides. (usually are, in modern "state of the art" ones) In order to minimize noise and stress on parts and whatnot, a good cam design arranges for all derivatives of all those curves to be equal at the point where they change over. As you will recall from your calculus classes, the first derivative is the rate of change in the valve's location; the 2nd derivative of its location is its acceleration; the 3rd derivative is "jerk", the rate of change in acceleration; the 4th, 5th, & 6th derivatives are called "snap crackle & pop" and are a little harder to relate any physical phenomenon to, but are still just as real. Given that mathematical constraint, it's REAL TOUGH to design a cam that snaps the valve open as fast as that .750" example up there, but stops short of .750" of full lift. You get the lift, even if you don't "need" it, just so that the parts are moved in a manner consistent with long life and all that.
Which is why it's fallacious (and "wrong" too) to pick a cam solely on the basis of the lift of "maxed out" head flow.