Your second point is wrong. Obviously it narrows specificity by half the room on average. I don't know why you'd argue against that obvious fact. It's basic math: If you have a set C that is the union of two sets A and B where A and B have the same cardinality, referring to "a C" gives you twice the possibilities than referring to either "an A" or "a B." So it's measurably twice as efficient to do the former. Since so many unrelated languages in the world ended up with such a system (or very close), it's reasonable to think that that efficiency was worth it. Since most of those systems are not much more specific, it's reasonable to think that being more specific wasn't worth it (one can always specify further using more words.)