Even if you level the reticle, which is what I do, if it's off kilter w.r.t. the turrets then the inputs from the turrets won't be the natural orthogonal vectors implied by the coordinate axes, i.e. the reticle.
Consider an extreme example, a 45 degree can't. You want to move bullet impact one unit vertically. A one unit change on a turret will then produce .707 units of vertical and .707 units of horizontal change.
If Leupold allows 3 degrees of cant, as was mentioned by another poster, the error will be negligible out to a few hundred yards. It becomes more pronounced at longer ranges. How much error is introduced by a canted reticle isn't the issue however. QA and customer satisfaction are.
As I posted earlier, for $1500, Leupold should rectify the problem. Time to chat with a company rep, not the service department. A personal approach is best. Phone them again, or if Leupold won't deal with the problem, drop by the store and tell them you would like to chat with their rep. The store may even offer to help.
