> when there is so much evidence that total productive work per week doesn't increase beyond around 40 hours of work for most knowledge worker jobs

I don't think this is true, though maybe you can find a source to prove it?

What I've often heard and read is that the percentage of work falls off after a certain amount of hours per week.

or put another way, working 80 hours may not give you a doubling of productivity over 40 hours but it will increase the amount of productive output.