Just came across an example where:

<phr type="n">√ƛʼécʼ-p=ap Račəméntən</phr>

got autophonemicized to:

<phr type="p" subtype="i">√ƛʼícʼ-p=ap ʔḥacmíntn</phr>

The capital R was correctly changed to ḥ, but where did the glottal stop come from? Did the "insert glottal stop before a word-initial vowel" rule somehow apply between the removal of the R and the addition of the ḥ?

Hmmm.

There are 15 other instances of word initial ʔḥ generated from R in the data. I can just fix them with find-and-replace, but I'm curious about how they were generated.