It tells you whether or not the machine can pass for a human in conversation. It is not limited to receptionist skills. And it is as real a scientific test as testing to see if a drug performs better than a placebo.

Let's say I told you that the head of a university philosophy department was actually a robot and, not believing me, you went and talked to her for an hour where you were free to discuss anything you wanted. If at the end of that hour you were convinced that she wasn't a robot and was human, would you then conclude that the professor was only fit for receptionist work? If so, that would seem to be an odd conclusion to arrive at. The professor's ability to have a discussion with you and convince you that she is human does not mean she is only fit to be a receptionist. The same applies if I told you that a plumber was a robot and you spent an hour talking to her. Again it would also be odd if you concluded that she was only fit for receptionist work. In fact you might conclude that the plumber (or the professor) might make a lousy receptionist, but that doesn't mean you'd think they were robots.

Or to look at it another way, if you talked to a brain damaged person and discovered that they couldn't have a normal conversation and that they couldn't respond in a way similar to most unbrain damaged people, would you conclude that you had learned nothing of interest about that person? Would you conclude that conversation provided no evidence on whether or not the brain damaged person could perform intellectually on the same level as an unbrain damaged person? I doubt you'd reach that conclusion.