That is one of the authors points -- no one has done self replication on a macro scale and nanotech just assumes self replication at the molecular level
Drexler's egotistical to be sure. It doesn't make him wrong
Also, self-replication is actually easier at the molecular level because all the parts are perfectly regular.
And ducks can fly with flexible wings.
And the Sun is a thermonuclear reactor.
You can find demonstrations of this type of things everywhere, that doesn't means that we can make them work with our materials, techniques and knowledge.
Nanotech seems to my like AI, always there, at just two steps... that's the type of problems that smell to vaporware from miles.
Now if only we could re-program women to make iPhones. The late-night trips to the gas station and gravel store because wife has another petroleum and sand-craving would totally be worth it.
The life if full of examples of things we know how to use but that we don't understand. Sometimes it's just impossible to control that things. Think in the stock market or economics. Chemist has been studying this type of things for centuries and you can see the result: there are chemical products everywhere you look.
What the post attack is the concept of creating nanomachines that doesn't degrade (Cosmic rays! oxidation! temperature!!) and do exactly what we want.
And about the "artificial DNA", it's something so simple... Really! the problem is in the emergent properties of DNA, proteins and environmental effects, this is where you loose control. Even nature has problems with that, life forms die, had cancer...