The evolution of P/F vs T/K raises important questions

[Jorge_Stolfi]

One simple fact may explain most of this. Plain p and f are almost never followed by e ... What does seem to hold is that where k/t take e, p/f take ch

Again, in my recollection the left context of p/f (distribution of letters that can precede p/f) is rather similar to that of k/t; but the right context is different from that of k/t -- but somewhat more like that of ke/te.  Which seems consistent with your observation of the increased frequency of "ch" there.

Hence my hunches that (1) p/f is a fancy form of either k/t or ke/te, and (2) which of these is the case depends on whether the horizontal arm has a hook or not (that is, the hook would be the fancy representation of the "e").

Unfortunately the "hooking" seems to be a continuous attribute rather than a binary one.  Although that ambiguity could be a "contribution" by the Scribe, or by the global Restorer...  

Anyway, AFAIK all digital transcriptions ignore the hook and encode only p/f.

All the best, --stolfi