What matters is the sequence of data, if any, that might be stored in or on such structures.
Watson's math simply applies to the probability / improbability of such data self-forming naturally (i.e. without intelligent intervention or aid).
So you want to consider DNA as a black box? I believe his argument gets even more tenuous at that point. While he does address the concept that there are some errors, I don't believe that he does enough on this tack. There is no experimental evidence that there aren't many more different combinations than he recommends, enough to blunt the odds that he calls impossible.
No evidence is no evidence. If there is no evidence, then you can't logically claim that Watson's odds are blunted.