It's measuring a random quantity in the bigger picture.
Retention percent doesn't matter as much as number of users retained, which leads to total revenue. But then you have to factor in the cost of acquisition. And the cost of infra to serve non-paying users.
My take on it: they took a multivariate analysis, plugged in the details of their business model, came up with a result, and then publish it as if it's advice applicable for all kinds of companies.
When a business is new, you don't want every client, you need the "right" type of client. The ones who don't fit (yet) can be a huge drag. After a more general PMF they're ok.
I think the article had this generalization but went too specific (for their case) and then lacked numbers.
This is why I always put typos in my spam emails, and make sure to always mention that Javascript is a good programming language at least once in all of my HN comments.
There are very few businesses that get away with this, of those - they are mostly entrenched monopolies that can get away with anything that doesn't cause regulatory reaction.