Because as you get a smaller and smaller skill set, that set of skills is not found in all of the populations.
So at each level we are looking at you would actually be eliminating some small schools from the equation at each level.
Back to the height idea.
If we said 6 Foot tall= sectional roster. It would be expected that all schools regardless of size would have 14 six foot tall kids.
If regional s= 6'5'' Then it is probably safe to assume that most schools have at least 1 kid walking the halls that is 6'5". You would naturally expect there to be more 6'5'' kids at a big school, but most schools would still meet this criteria.
Semi-state- a bigger jump so lets say 6'10, at this point most big schools would still have a 6'10'' kid, or maybe a few walking around. Several small schools would too, but now you would start to eliminate several small schools entirely, because they simply do not have a 6'10'' kid walking around their population. The pool of schools at this point is lessened, this is why you would see the drop off. Now instead of 100% of the 13% we are dealing with 75% of that 13%. At the same time we are probably dealing with lets say 90% of the 24% (2A schools) and 100% of the 63% (big schools).
State= 7' Now we would see that not even all big schools would have a 7' kid walking around. Small schools it would be very rare, but there would still be some schools with a 7' kid. So now lets say were are looking at a populations of 30 % of the small schools, 45% of the medium schools, and 75% of the big schools.
I could keep going on with this, but the point is that as the skill set becomes smaller, the % of schools that had a kid with that skill set to begin with gets lower, thus why we see the small disparity in qualifiers, and the larger on in placers.