I don't believe this to be correct.
If he were right, then applying a curve before RAW conversion would be exactly the same as applying it after RAW conversion and it has been shown that the application of a curve to convert the linear information captured by the sensor into a logarithmic space is an entropic, non reversible, process where some information is potentially lost.
By doing 2 conversions with different curves, it is possible to keep in the 2 converted images the most useful amount of local information that can hence be overlayed into a single image containing more information than a single conversion could.
In other words, if local curve application were possible in the RAW converter, then there would indeed be no advantage in overlaying images coming from 2 conversions, but this is currently not possible.
Someone with more knowledge correct me please.