Independent contrasts were estimated using the primate phylogeny presented in Purvis [19] with all branch lengths set to 1 (Text S1).
http://plosone.org/article/info:doi/10.1371/journal.pone.0000062#article1.body1.sec4.sec1.p2

The tests for phylogenetic independence were done using branch lengths set to 1 as well? If so, it's rather surprising that a phylogenetic signal was found. I would suggest using branch lengths based on other data, such as cytb sequences, and re-run the tests for phylogenetic signal with those new branch lengths to assess whether the data is not an artifact of the method used.