Comment on “Mutually unbiased bases, orthogonal Latin squares, and hidden-variable models”

In a recent article Paterek, Dakić, and Brukner [Phys. Rev. A 79, 012109 (2009)] show an algorithm for generating mutually unbiased bases from sets of orthogonal Latin squares. They claim that this algorithm works for every set of orthogonal Latin squares. We show that the algorithm only works for particular sets of orthogonal Latin squares. Furthermore, the algorithm is a more readable version of work previously published [Phys. Rev. A 70, 062101 (2004)].