If I had to guess, it's that these studies aren't sensitive to the outliers. Imagine telling e.g. Elon Musk "multitasking and long hours are bad for your productivity".
I've been keeping very detailed data on my personal productivity for years, while taking all of the consensus wisdom ("long hours = bad", "multitasking = bad") to heart. What I've instead found is that long hours (literal 14+ hour days) makes me more consistently productive.
Multitasking also has a strong synergistic component to it that I suspect doesn't show up in these narrow studies.* If multitasking across different projects leads to insights that shave years off of your respective project plans, but otherwise makes you slower locally, it's still more than worth it. Again using Elon as an example: Tesla and SpaceX have been made mutually stronger by insights being cross applied the two enterprises.
* To be fair I'm speaking more of multitasking at the project level, not at the minute-to-minute level.