Craig Melchert, the Diebold Professor of Indo-European at UCLA, has kindly forwarded a list of addenda and corrigenda which I will be posting beginning with this one.
On pg. 35 I quote Craig's famous 1987 article on the three-fold treatment of the voiceless velars in Luvian, but this needs modification. Contrary to the claim of that article there is no good evidence for the affrication of the voiceless palatovelar before a back vowel. In particular, the supposed case of *-(i)k'o- > Luv. -(i)zza- ~ Lyc. -(i)s(e)- does not stand up. The Luvian suffix -(i)zza- probably goes back to *-tyeh2 and the Lycian suffix -(i)s(e)-, which makes abstracts and place names, continues *-sh2o-. Instead it now seems likely that there was a conditioned fronting of the voiceless palatovelar before front vowel (Luv. ziyari ‘lies’ (C) ~ Lyc. sijẽni), yod (wazi- ‘request’ (H) < *wek'-ye/o-), w (azu(wa)- ‘horse’ (H)< *ék'wo-) and syllabic sonorants (zurnid- ‘horn’ (H) < *k'r̥ng-id-, zanta 'down' < *k'm̥t-). Thus Luvian and Lycian are like mirror images of Albanian with palatalization differential affecting the palatovelars. We still have a ternary reflex, just not an unconditioned one.