I think it has actually more to do with "But they also needed to absorb light at different rates to buffer against the external noise caused by swings in light intensity."
To create this buffering one would ideally have the the least proximal wavelengths, which will presumably be least similar in downstream rate of conversion/transportation and therefore smooth energy flow from the flickering, but which are still at near peak intensity. In this case that's blue and red visible light. The article notes they are at the steepest part of the intensity curve which is just a proxy definition for the above criteria.
This of course still assumes that their model of generalised networks is analogous to the real biological cascades - otherwise could just be another coinciding superficiality rather than evolutionary reality (https://news.ycombinator.com/item?id=33050912). Quick skim through the source paper on SciHub leaves me sceptical.
This of course still assumes that their model of generalised networks is analogous to the real biological cascades - otherwise could just be another coinciding superficiality rather than evolutionary reality (https://news.ycombinator.com/item?id=33050912). Quick skim through the source paper on SciHub leaves me sceptical.