Well, CERA, as you mentioned, does show a very significant differential between Leon and Vaz/Swihart with Sale & Price. The sample sizes are pretty large.
Career
Sale:
2.50 with Leon (1273 PAs)
4.50 with Vaz (279 PAs)
3.23 everyone else but Leon over his whole career
Price:
2.83 with Leon (769 PAs)
4.32 with Vaz/Swihart (1234 PAs)
3.31 with everyone else over his whole career
Do we "know" these numbers are all because of Leon? No, of course not, but do we know a career .300 hitter is a better hitter than a .295 one, since so many factors go into how a hit becomes a hit and is ruled a hit or an error?