To put it differently:

The analysis in the article doesn't normalize for the frequency of a certain founder-number bucket. What if 99% of all start-ups have single founders? Then of course most of the companies that exit have a single founder, even if single founders perform much worse.