> within the NHS recommended limit

Over the NHS recommended limit is better than zero caffeine for everyone. If their limit is correct is in question

Whether "those with severe mental illness" get more benefit seems unlikely biologically. But like everyone coffee is good for you.

The only point of research like this, since we know coffee is good, is finding the mechanisms. But it's highly open to p-hacking/experimental error, which is how universities work now. You should default to this is citation farming.