Yes, it does seem to be complicated and inconsistent. But personal followers and publication followers are not equal. If a publication has 1,000 followers, it's safe to assume that many of those will also be writers who are submitting articles to the publication as well. The more followers, the more competition too. But with personal followers, you're not competing against other writers in that way.

Then there are other factors, such as the topic you've written about, the tags you've used, how long the article is, and what time of day or week it was published. A properly scientific test would not be possible without publishing the exact same article in a publication and self published, and both getting published at exactly the same time. And you'd have to do that 100 times for it to really provide any useful data, by which point Medium would have suspended your account for duplicate content 😆