I agree with your overall point, in that the number of papers published != the amount of 'science' being performed, however that sort of metric might not be such a bad way to gauge the output of a shared resource like a telescope. For the sake of argument, if the telescope time is split up between 100 different groups, each of which submitted a proposal for which the end goal is a publication, is it not that unreasonable to assume that the publication would at least be available as a pre-print say six months after the group's observation time has ended?