Thanks for neatly explaining my scepticism of laboratory analogues and trying to draw conclusions from them. They're basically poor-man's computational simulations, with more errors because you've gone through another layer of model approximation:

    physical system -> theoretical description -> physical system

instead of

    physical system -> theoretical description

followed by solving it with a computer.

It's not surprising that when you set out to simulate a black hole you end up with a black hole! The main reason I could accept them being useful for is when some emergent aspect of the theory is very expensive to simulate.