The part of your argument that leaves me unconvinced is your use of the median income as a benchmark. The article referenced (on yahoo) mentions the high percentage of foreign nationals who obtain PhD's and Master's degrees at US Universities.
So you're comparing someone who was talented enough in high school to hit the ground running with Calculus at university. This student then sleeps very few hours struggling with advanced calculus, differential equations, stochastic processes, nonlinear optimization, compiler design, chemistry, and physics. He scores well enough on the GRE (and perhaps the subject test) to get into a PhD program, and makes it through 6 years + a dissertation.
You compare that to the median. Seriously?
Meanwhile, his fellow lawyers in training breeze through a Poly Sci major. It's tough to get into a top law school, but no harder than getting into a top engineering program (because you don't have to compete with foreign nationals to get into law school in the US on anything near the scale you do in engineering). 3 years of law school or 2 of an MBA program are no picnic, but virtually all students at these programs obtain their degress (which is hardly the case at PhD programs, I know this from personal experience at Berkeley).
Salaries would have to be as high or higher than Law/MBA to convince Americans to do this. Instead, engineering salaries can lag way behind, because it's easy to get foreigners, especially from countries without much economic opportunity, to sign up for this as a way of gaining a path toward US residency and citizenship.
While salaries might not be actively lowered, they almost certainly place a limit on wage growth. When this happens over decades (and wages aren't similarly controlled in other fields where Law and MBA students are hired), you'll start to see a huge wage differential.
To me, this has clearly happened - and now Americans have no interest in signing up for such an intensely rigorous path with rewards that are, quite frankly, much lower than in other professions.