As you say: Sorry, but no. I don't buy it.

This is the kind of thing you want data on, not hunches. Courts are not an ideal setting for determining such things, but at least they should be trying to make the best unbiased determination possible, keeping in mind the correct study results.

Basing a default attitude towards custody upon a hunch about the partitioning of narcissists vs. caring parents is a pretty bad idea. (Edit: this goes for repeating someone else's hunch or claim about what the data shows. You need actual data, actually presented with sufficient description of method to understand whether it should be believed.)

Otherwise, good article! There are far too many places where what looks like it might be causal is actually a selection effect (e.g. "watching TikTok reduces your chance of heart disease").