I think the assumption of the author is that the 3 hours he does work are best-case productive hours. If you work 40+ hours a week, you still have to account for all those staring-at-your screen hours, that are not contributing towards meeting the deadline.
Somewhat related: I found it is a lot easier (and natural) to estimate projects in terms of 'ideal' hours. And then just assume there are only 20 to 25 of those a week. That quite closely matches the 21 hours of the original post.
This is somewhat different but, back when I did consulting work that often involved writing in some form, we tended to think in terms of the work that we could do in a "production day" of about 4 hours. The thinking was that, on a given day, we'd spend about half on project-based work and about half on the phone with clients and others, doing miscellaneous research and "research," internal discussions, and various administrivia. Sometimes we'd buckle down more to get something done and other times we were at conferences and the like--but we found this a good rule-of-thumb.
Somewhat related: I found it is a lot easier (and natural) to estimate projects in terms of 'ideal' hours. And then just assume there are only 20 to 25 of those a week. That quite closely matches the 21 hours of the original post.