The default behavior of gregoriotex could be smarter, by detecting the note on the right of the dot, and centering the dot between the two noteheads at K height, at the expense of complication and/or the risk of breaking other cases where it works "well enough". Also, I argue that the Solesmes typesetting of clivis+quilisma-pes is not very good, since it uses a vertical bar to join the first two notes, that is artificially made "diagonal" to fit the dot.
In any case, I firmly believe that using the mora dot in this specific instance is plain wrong and the note should be episemated instead: k_i!jwk . If someone has insight as to why Gajard & co did not settle on the horizontal episema for this specific case, I'm interested.
Matthias