A newer paper from 2016 confirms the expectations that in mainland Croatia Z280 prevails. If you check ''Genetic heritage of Croatians in the Southeastern European gene pool—Y chromosome analysis of the Croatian continental and Island population'' (Jelena Šarac 2016) you see M558 described as a major haplogroup among Croats (19%) in the intro. In the same publication R1a in Croatia mainland consists of: 21-M558, 3-M458, 5-Z282. In Dubrovnik and elsewhere M558 is also described prevailing.

The 2016 publication shows 88 sampled from Zagreb, Zabok, Donji Miholjac, Pazin, Delnice. Although not 108 again, these are appearently the same samples (excl 20 from Dubrovnik) derived from "Pericic et al. High-Resolution Phylogenetic Analysis of Southeastern Europe Traces Major Episodes of Paternal Gene Flow Among Slavic Populatiions" which describes 108 sampled from Mainland Croatia from Zagreb, Zabok, Donji Miholjac, Pazin Delnice and Dubrovnik. My conclusion is that data from Underhill 2014 is somehow mistaken, there is no other explanation.

As you said, Serbian, Bosnian and Croatian must be basically homogeneous in terms of R1a, this is more plausible suggestion if we consider the linguistic category, as the Serbo-Croatian language quite resembles Ukrainian. M458 is only significant in Bulgarians (and Greeks) in the south, whose language is descended from the Polish tribes.