From my understanding, the nDNA is probably the more accurate one, as Rican makes the decision to keep amatitlania split from cryptoheros based on the fact that the hypsophrys/neetroplus lineage sat between them, and the fact that the latter two were distinct enough to be kept separate from cryptoheros and amatitlania. This "split" is only seen in the nDNA cladogram.
Among other things, without reading further into it, a few obvious "flaws" can be seen in the mtDNA cladogram, such as the fact that it implies bussingius is the sister genus to parachromis, the fact that myrnae is considered the sister species to kanna (outside bussingius), and the fact that mayaheros is shown as split in half (m. urophthalmus being shown as closer to p. splendida and m. beani being shown as closer to istlanus, which is true, but does not hold up when current descriptions are considered). Flaws as in things that don't hold up when compared to current taxonomy, not as in "they did something wrong when they made this".
Going off of the genera they erected in this paper I would consider it safe to assume the nDNA one is more accurate.
With regard to the kanna myrnae thing in the mtDNA one, I saw the same thing in the Bagley paper lumping amatitlania. The closeness of kanna and myrnae in that paper was explained as being the result of a possible hybridization event that occurred somewhere in the ancestry of the myrnae sample(s).
