Well, it's quite unlikely that 40 hours/week is exactly the right value. I'd expect that what's going on involves researchers comparing the cultural default to a grab-bag of longer hours, probably with fairly coarse granularity, and concluding that the cultural default works better even though it might not be an absolute optimal.

There's also cultural factors to take into account, both local to the company and general to the society. If we've habituated ourselves to thinking that 40 hours/week is normal for people in general, it wouldn't be surprising to me if working longer hours acted as a stressor purely by comparison with others. Similarly, among companies, expecting employees to work longer hours than the default would probably correlate with putting high pressure on them in other ways, and this would probably be very hard to untangle from the productivity statistics.