Any Insights About Annual Stamp Production Versus Mail Volume Processed?

I'm working on a research paper which involves comparing the annual amount of stamp production to the volume of mail processed - for 1921-1939 in this instance, but my questions below are more general.

For the US, for example, I found that the relation of annual stamp production ("ordinary postage stamps", not revenue stamps) to mail processed was between 75-80% annually (and that this shortfall of 15-20% was made up by stampless mail, like metered mail). My question here is, is that kind of statistic possible? Shouldn't there be any amount of "overproduction" of stamps which would account for leftover unused stamps? or is overproduction so small in the scheme of things that it wouldn't change the relation by more than a fraction of a percent?

On the heels of that question I have another: the sources I used for US stamps cited both stamps produced as well as stamp sheets printed. I was under the impression that an average US stamp sheet of this period would have at least 200 stamps (in some cases also 400). But when I compare the statistics cited it seems the ratio is more like 105 stamps (i.e. 100) to a "sheet". Does that make sense?

The sources I used were various editions of the "Annual Report of the Postmaster General" and editions of the "Treasury Department Appropriation Bill" - that latter using references to "sheets". Both are available freely on the internet.

You may find some indication if you can find assessments by the post office of unused stamps in circulation for the purposes of determining financial liability. I seem to remember that the GB post office carried out surveys of this kind in the past, but I don't know howfrequently or how accurately, or whether the information is readily accessible.

The era you cite has a huge shift from stamps to meters. I seem to recall that by the early to mid-1950s, metered mail surpassed stamped mail in the U.S.

Sheet/pane: I suspect in the various reports you are citing, the term "sheet" is likely misused by the authors and refers to what collectors would call a retail sales unit of a "pane", which is typically going to be 100 for most U.S. definitives and 50 for most commemoratives, not to mention booklets and coils.

My gut feeling is that there are too many variables in the equation of rates/classes and methods of payments to draw more than broad generalities.