The question that most people are interested in is
P(Success | Number of Founders)
but what the article has answered is
P(Number of Founders | Success)
for a couple of different metrics of "success".
They are not the same question! In particular, the average number of founders for a successful startup might be low, but the chances could still be better with more founders.
All startups are successful if you ignore the ones which fail. True, but not very interesting.
Not excluding the failures is the point of the exercise. One cannot choose to succeed, one can only choose the number of founders. The question is how many founders leads to the greatest probability of success.
Although approximately 50% of successes have one founder, it does not follow that approximately 50% of startups with one founder succeed. If 70% of startups have one founder, then those startups are under-represented amongst the winners. OTOH, if 20% of startups have two founders then, with 30% of winners having two founders, they are over-represented.
If the numbers I just made up were accurate, then it would be better to have two founders than one, even though most successful startups have one.
They are not the same question! In particular, the average number of founders for a successful startup might be low, but the chances could still be better with more founders.