On Eldamo's Sindarin Phonetics, it is assumed that CjV > CiV happened before the first stage of i-affection. Are we entirely sure this is not the other way around? I-raising is effectively the opposite of a-affection, but a-affection is hardly undone in any word that ended in -ja in Old-Sindarin.
If the suffix was already -ia before i-raising happened, wouldn't we expect delia to be dylia (dulja > dolja > dolia > dulia > dylia), nedia- to be nydia-, and gwain to be gwîn (gwinja > gwenja > gwenia > gwinia > gwini > gwin > gwîn), and mein to be mîn, feir and fair to be fîr, dail to be dîl, etc.?
We could assume that -ia did not instantly cause i-affection - maybe i was still too j-like, it was a gradual process after all, or maybe a-affection was still an active enough process to prevent this - but wouldn't it be less far-fetched to assume that -Cja > -Cia after the first and before the second stage of i-affection?
For the sake of completeness, I found one counterexample, tiria- is not teria-, but that might be because it was/is tíria-, and this is the only counterexample I could find.