Why not positional variation?

[Mauro]

My reading of this result is that in and iin words must be the same. The author's mind is sometimes in single stroke mode, and at other times in double stroke mode.

I think this is improbable, because the frequency of -in or -iin endings is different from one word type to another.

Just for example (this happens all over the place): there are (RF1a-n text) 834 'daiin' and 202 'dain', with a ratio of about 4.1:1; 'aiin' and 'ain' behave similarly, with 529 and 123 occurences, about 4.3:1, but 'okaiin' and 'okain', with 205 and 126 occurrences have a ratio of 1.6:1, while 'qokaiin' and 'qokain' have 259 and 274 occurrences for a ratio of 0.95:1.

If -in and -iin were really the same I would expect rather similar ratios among all words with those endings (with the caveat that the data above are for the whole text, I cannot exclude the ratios become all similar considering the VMS section by section, or Currier language by Currier language etc.).

P.S.: the wildly variable ratios are what makes it difficult to model the VMS with a Markov chain (I tried once to do it, but it needed so many nodes to accomodate all the 'X-aiin/X-ain frequencies are different from Y-aiin/Y-ain ones' cases that I gave up, it was getting absurd). Always with the same caveat as before!