You're right, as far as it goes. To be optimal, it might be good to pick from a certain class of common username (different on each site) and use that. For example, maybe I could be "cardinals59" on one site and "I|1II||1II" on another site.
However, that would offer limited anonymity benefit. With the full firehose of Reddit and HN data, you could probably correlate my accounts on each site to a moderate degree of confidence just by when I post. But, I'm not too worried about that, as long as a third party can't easily link that to my real life identity. And you can see by my post history that even in that case, I'm not trying to make it impossible. Just difficult for the casual harasser. With access to the biographical information I've posted on this account, you could probably guess my employer. With access to that entity's internal databases, you could probably narrow me down to one of about 5k people. With access to other information I've posted and public records, you could probably narrow down to one of hundreds of people. Maybe if you had a lot of skill you could do more.
But, why? It's all about risk management. As long as it's hard enough to do that it's lower than other risks I face, I'm fine. If I were higher profile (e.g. if I were an SVP at my employer, or were likely to become one someday), I might be even more discreet. It would be most sensible not to post at all in that case. But I'll probably never be that important, and that's how I like it. As it is, the risk of my online identity being used to hurt me is (I estimate) lower than the risk of being killed in a traffic accident. I don't worry too much about the latter risk, so I really shouldn't worry much about the former.
For me, it's good enough that if you Google for my full name, you don't see me. Even if you google my name + employer, or my name + hometown or current location, you can't find any information on me. That's the extent of the threat model I'm trying to actively defend.