They only tested in the range from 350 to 380nm [1]. With wave lengths under 370nm the efficiencies are quite high; at 380nm it falls to 33,6%. That could either mean that the catalyst works well with light at small wavelengths or that it only works at this specific range.

[1] https://www.nature.com/articles/s41586-020-2278-9

It's probably the same as with solar cells, where you need to match portions of the spectrum with appropriately doped silicon.