The second issue is the free space pathloss (see https://en.wikipedia.org/wiki/Free-space_path_loss ) you can see by the formula that the loss goes up exponentially by not only the distance but also the frequency, which is why the wireless carriers are so interested in the 600Mhz spectrum, you need to build fewer towers using 600Mhz than say 3.5Ghz.
The sudden interest in mm-wave (e.g. 28Ghz) for 5G is because the spectrum is much wider, that is the channels can be very wide allowing more data to be encoded, the trade off is that you start losing the NLoS capability, or it becomes much more difficult.
With light, all those problems are orders of magnitude more difficult to overcome. Also, there are regulations around the level of emissions of light for safety reasons, if a laser is turned up too high it can harm a persons eyes if they happen to look at it, so although the spectrum itself is not regulated, the amount of power you can emit is. That makes it very difficult to implement an FSO system that can safely transmit more than one or two kilometers, but if you get fog, rain or snow, that distance quickly becomes 100's of meters at best.
They are correct that the best way to overcome this issue is by developing sensors with much better sensitivity, by making them larger you can theoretically collect more photons, like a telescope. Since it is unlikely you can overcome the safety issue of increasing the transmission power level, it is smart to focus on the receive end and improve the sensitivity. But that does not help the LoS problem.
But a very interesting development nevertheless.