edit: actually, I'm wrong, see below

You're right, I believe. I don't have time to investigate, but if the claim is 90% efficiency w.r.t. total radiation input, it's certainly wrong. Doing so would violate the 2nd law. For example, room temperature thermal radiation has significant emmitance in the radio range of the spectrum. Needless to say nobody can build radio antennas to turn this energy into work.

If it is 90% of the thermodynamic limit, then maybe. But then I think multijunction cells are already significantly close to the thermodynamic limit (much better than 22% claimed).

https://en.wikipedia.org/wiki/Carnot%27s_theorem_(thermodyna...

(obs1: substituting Tsun as Th and Tearth as Tc left as an exercise to the reader ;) )

(obs2: try changing Tc to CMB temperature! )