There is a real problem with 'objective' HCI studies which surprisingly few usability experts address:
The majority of these studies are done specifically by finding people who've never used the environment before, because they want to avoid users having preconceptions about the system. This seems fair, but unfortunately the result is a tendency to optimise things for the first few hours of use - or perhaps the first few days/weeks.
In reality nobody is a newbie forever, so interfaces which have great usability initially (ie 'easy to use without learning') are often less usable in practice (ie poor learnability/flexibility).
I'm sure it's *possible* to design an interface which is welcoming to first-time users without being restraining to long-time users, but thus far the free software desktop has a poor track record in this department.
BTW I think desktop environments have a particularly difficult problem here. With most software, you know roughly what the user is trying to do - after all, you know what your software is designed for. A desktop environment is different; it needs to support an enormous range of use cases and can't safely make many assumptions about the user or their goals.