I'm not sure if you wrote it or if someone else at CB Insights wrote it but it is interesting. When you looked at the data, did you considered the following two possibilities:

1) The founders who raised from VCs for their seed understood how to raise money from VCs and were much more likely to be able to repeat the process.

2) The VCs had better deal flow and were able to finance companies with higher probabilities of raising additional rounds.

I spent a few years working with an angel investment presentation group at my university. I noticed that the deal flow was primarily companies that could have a $20m-$50m exit, but were never going to be mid caps or large caps and that exit value is largely ignored by VCs. The more angel money that funds those, the less opportunity for follow on rounds for the group financed by angels.

Additionally, the situation your data describes doesn't really fit the Y Combinator example since qualitatively they are very different. The VCs that invest in Y Combinator companies at the seed stage do so in batches without analysis. That changes over the course of time, after they have made the investment as they get to see progress. The fact that they make the original investment blind, then later make the second investment with better information causes your data to not be applicable to the situation.

Now, before you think that I am saying that the signaling is an issue, understand that I do not know if it is or isn't. I was simply pointing out what I believe the author meant.