No, it doesn't have to be all-or-nothing. The flaw in the Graham's argument is that he argues we should prefer "natural" without first proving "natural" means "good". Bubonic plague is very natural, but not so good, right? That doesn't have to be all or nothing, but each time somebody says "do X not Y because X is more natural", we ought to ask "but does natural here mean good?". Sometimes it does - eating an apple is probably better than ingesting a bunch of high-inducing refined sugar - but sometimes it doesn't. Antibiotics are not natural, but preferable to being eaten alive by bacteria. So this needs to be established on case by case basis - is "natural" good in this case? And Graham doesn't seem to bother to do that.
David Hume pointed this out in 1740, and his advise still applies:
A reason should be given, for what seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely different from it. But as authors do not commonly use this precaution, I shall presume to recommend it to the readers.
I think paul graham is biased to always tell people to make startups, but what hes saying resonates to me in that it is a spectrum. You dont have to. Ake a startup, but you can work for a company of 20 people and feel thats its alot more liberating and feel a much stronger sense of purpose. He even concedes that near the end