To the contrary, it presumes that the right edge of the curves for both applicant pools do _not_ connect at n - ~100. Given the size difference between cohorts, I argue the curves would never meet at all. If the sizes were identical or close, then you could argue I'm using the fallacy.
I had zero interest in Caltech for reasons I probably couldn't tell you years later from when I was a young man. Ended up at a different engineering school somewhat by default and it was good for me. (Never even went on campus before I started.) No regrets.
Your second guess is correct.