Me thinks they are referring to the duration you can integrate the signal before reading it - which by itself is very useful.
*This is for "non-coherent" systems. Every application mentioned in the article supports this. Nothing in the article indicates otherwise.
There is no color information when you are converting photons to electrons, right?
it also uses 10 times less energy
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Is this to say that it creates energy?
Not necessarily. The ultimate limit that one can push a photodetector to is, in the final analysis, determined by "noise", not conversion efficiency. If you can reduce the "noise" to 1/1000 of its previous level, you can increase the amplification applied to the signal by that same factor. Just increasing the integration time won't help if that integration time is also accumulating "noise".
Which is why many astronomical cameras are liquid nitrogen cooled.