They're not meant to be useful from the perspective of cryptography, in the same way as learning textbook RSA is not useful in that sense. They're useful in the sense that people want to have some faint clue of what's happening.
Maybe it's just a difference in how I look at it, but knowing what group law is doing is not a useful clue to me.
From being able to apply group law you couldn't predict the properties or (most of) the vulnerabilities. (or even performance, since these things don't generally cover projective coordinates).
At least the way I see it is if you asked about how a spell checker worked, and then I set out and explain how a digital adder circuit works from the gate level. You can't build a spell checker without adders yet you wouldn't be informed. :)
This sort of thing also strikes me as the sort of thing that more "feeling of understanding" without creating much understanding. I seldom think it's good to do that, though it isn't always harmful.
But learning styles is certainly something that differs a lot between people, so I certainly don't think my experience is universal by any means.
I can say that personally learning to ignore the machinery made my understanding grow 100x faster and directly resulted in constructing more than a few publishable (and widely deployed) designs.
