I think that yards allowed per coverage snaps in combination with TDs and interceptions is a pretty good indicator of a cornerback's performance.
I'm highly suspicious of yards-per-coverage-snap data.
Passing offenses are "right handed" (or "left handed" in the rare cases of lefty QBs; the rest of these observations will assume a righty QB). I can't recall if I saw Rodgers' intermediate-to-deep 2014 throws charted in a link posted here or elsewhere, but even
his throw frequency and completion rate on those passes is predominantly to the right side. The difference was quite striking, in fact. It's just plain more difficult to turn and throw left in a mechanically sound position, i.e., not across the body, than it is to the right side...there's simply fewer degrees of turn going right.
Further, the QB's vision dropping back from center is to the right side. They don't call the left side the "blind side" for nothing; if the QB is blind to the rusher, he's bound to be at least somewhat blind to the routes on that side as well.
There's also the matter of called roll-outs and ad lib plays when bailing the pocket...those are predominantly right-sided specifically to avoid having to throw across the body. Some are better than others at going left. Rodgers may be the best throwing across his body, but there's bound to be a drop off in range and accuracy regardless. With lesser QBs the difference is more pronounced. When watching QB coaching sessions with high draft prospects ((Mariucci, Chucky, et. al.) you see them hammering away on footwork and mechanics on the left-side throws.
There's no getting around these left-right factors; the closest you can get to mitigating those factors is using shotgun, where the QB is not back-pedaling facing right.
Consequently, the default position in the NFL is put the #1 receiver on the right side, with the defense sticking the best coverage corner opposite that receiver at left cornerback. Certainly, there are plays where the offense tries to exploit a match up and switch the #1 receiver to the left side against the weaker corner. And some defenses switch sides with their cornerbacks to counter. But the default is strength-against-strength on the right side.
To illustrate, this week's reports from the underwear workouts have Shields working left cornerback (offense's right side) with the first team, taking over Williams' spot.
Then there's the matter of comparing nickel corners to cover corners. Nickels do a lot of duty against slot receivers and tight ends in short routes and/or in the crowded middle of the field where there is frequently help, minimizing yards even if the coverage is blown. In a yards-per-coverage snap comparison, nickel corners clearly need to be evaluated separately from cover corners.
More subtly, cover corners who are fixed at left corner, or who follow the #1 all over the field need to be evaluated separately from those that are fixed at right corner or who follow the #2.
Even more subtly perhaps, corners that play in a defense that dishes a high frequency of two-high safety are afforded more protection from long yardage plays than a defense that plays a lot of single-high safety.
And then there's the difficulties in accounting for the help that a good 4 man pass rush provides. How do you compare Buffalo's corners, where the defense will go entire games in nickel and dropping 7 while getting pressure with 4, to a team with a lousy pass rushing front that relies on blitzes? Or what if you have a bad free safety, as with M.D. Jennings, where the corner is burned and the end zone yapping begins over a blown top-side assignment?
Then there's the weak link factor...if there's a big fall-off from the #1 to the #2 corner, opponents will be inclined to go after the weakness. In a way, a #1 corner will look better in yards-per-coverage snap if his compatriot on the the other side stinks. That is, after all, what a good deal of game prep goes into identifying, or as the euphemism goes, "finding favorable match ups." Under this scenario, the offense is likely to stray more frequently from the strength-against-strength default.
All in all, evaluating corner performance in statistical isolation is a very tricky business.