I have written above a while ago (post #65) that the Amaurornis akool mt-genome was wrong, because large parts of it are near/fully identical to Gallinula chloropus. Additionally, the remaining parts of this mt-genome (that do not look like Gallinula) do not, actually, look like other available sequences of A. akool either. They look like sequences of A. phoenicurus. (Three cox1 and two cytb from GenBank (from [
Yang et al 2010] and [
Ruan et al 2012]) and, as you note, the analysis of cytb data by [
Slikas, Olson & Fleischer 2002], all indicate congruently that akool belongs in Zapornia, not in Amaurornis, where it ended up in Boast et al's analysis.) This mt-genome is clearly chimeric, resulting from a mixture of the mtDNA of (at least) two taxa, neither of which appears to have been genuine Zapornia akool. Unfortunately, this mitochondrial genome was included in Boast et al's data set.
Chimeric sequences, in cases where the two involved taxa are also included in the analysis as well (as here), are attracted simultaneously towards two different parts of the tree; additionally,
through them, these two portions of the tree become unduly attracted towards each other:
the presence of a chimeric sequence in a data set will affect the tree in ways that go far beyond a simple wrong position of the taxon the chimera are supposed to represent. If one of the taxa is dominant in the mixture, the chimera may remain associated to the group this taxon belongs to. But the attraction towards the other involved taxon (which lies outside the group) will then often result in it assuming an unduly basal position in this group. If the topology within this group remains nevertheless 'correct', the whole group will then end up completely mis-oriented. A usual 'symptom' of a mis-oriented group is that the branches in the group become shorter and shorter as they become more basal -- this is exactly what you see in Boast et al's Fig. 3: the branch to the chimeric "Zapornia akool" is shorter (i.e., the tip ends up positioned more to the left) than that to Amaurornis phoenicurus, which in turn is shorter than that to Amaurornis moluccana/olivacea, which in turn is shorter than that to Gallicrex. Don't misunderstand me -- successively shorter branches are no proof that a tree is wrong; heterotachy
can happen. But, when you see this type of pattern in a clade where the basal-most sequence, at the end of the shortest branch, is demonstrably chimeric, you should definitely not bet a single coin on the apparent relationships within this clade. Without the chimeric "akool" sequence, it is quite likely that the group would readily re-orient with Gallicrex sister to the three (other) Amaurornis.
I see nothing suggesting problems in Slikas et al' paper, although they never released their sequences (deposition in GenBank was not yet a standard procedure back then), thus it is impossible to check how they would behave in an analysis that would also include more recent data. The topology they found seems in any case compatible with what more recent analyses have indicated.
(More recent does not necessarily mean more reliable. Neither does a bigger data set, if there are problems in it, by the way. In my view, reliability increases mainly as a result of congruence between the findings of multiple studies, old and new, using independent data. (The latter point sometimes becoming a problem these days, as, more and more, studies are in fact based for a large part on the same data as earlier works. Of course, the congruence between the trees of Boast et al 2019, and those of [
Gong et al. 2017] (who produced the A. akool mt-genome that Boast et al re-used) is no indication of reliability. The trees congruently show Gallicrex embedded in Amaurornis, as a direct result of them being constructed from non-independant data sets. Both analyses are in direct conflict with, e.g., [
Garcia-R et al 2014], where akool is in what the authors called "Porzana" (i.e., Zapornia) and Gallicrex not nested in Amaurornis at all.))