I agree with you that p/f are not the same as ke/te. Some time ago I tested this hypothesis by measuring the (root-mean-square, Euclidean) distances of the distributions of the preceding and of the following character for the glyphs 'p', 'f', 'ke' and 'te' and they turned out to be quite dissimilar. 'ke' and 'te' were instead most similar to... themselves. The post is You are not allowed to view links. Register or Login to view.
I explored a little more the behaviour of 'k' and 't' ('p' and 'f' are rare glyphs and I'm not sure if statistics on those would be reliable), which are quite similar (by the above metric) among them: it turned out 'ok' and 'ot' are even more close than 'k', 't' alone, and removing 'ok'/'ot' removes also most of the similarity between 'k' and 't'.
I explored a little more the behaviour of 'k' and 't' ('p' and 'f' are rare glyphs and I'm not sure if statistics on those would be reliable), which are quite similar (by the above metric) among them: it turned out 'ok' and 'ot' are even more close than 'k', 't' alone, and removing 'ok'/'ot' removes also most of the similarity between 'k' and 't'.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)