Maxwell-Boltzmann statistics, and the field of statistical mechanics in general, work quite well with quantized systems. As an example, if you look at Boltzmann's definition of entropy: S = k ln W, where W is the possible number of microstates that can contribute to the system, you can see how statistical mechanics does a good job of handling quantized energy levels. Likewise, the Maxwell-Boltzman distribution does a fine job of describing the population distribution of an equilibrium ensemble of molecules / atoms / whatever with discrete quantized energy levels. The critical term here is equilibrium. If the system is not in equilibrium, such as a laser, then one can argue that it's temperature (at least for the degrees of freedom where there's a population inversion) is not well defined.
The thing that makes the Science paper really interesting is that the negative temperature is observed in the motional degrees of freedom where you normally think about a continuum of energies, and where you seldom have the necessary isolation from other degrees of freedom to prepare such exotic states. The key here is that Bose-Einstein condensate have coherent, quantized motional degrees of freedom that are highly decoupled from the rest of the universe.