If you look at the pictures closely and visualize where the lever's pivot point is for each of the two cap irons, you can see that the bad one's pivot point is further "back" along the long axis of the lever. That wouldn't impact its flushness on the finished side, though it would impact the gap between the base (non-toggle, cam end) of the lever and the cap.
Yep, that's the "file fix" that I alluded to in post #2. It's not without downside though. If you look at the right picture you'll see that he'll have to file the entire end of the lever off, and that will create a larger (cosmetic) gap between the base of the lever and the cap when the lever is in engaged/flush position.
The more I think about this the more I think L-N may have botched a design change or mold replacement here. From personal experience as a product designer it happens to even the best teams sometimes (though far less often than to not-best teams). I'm no longer convinced that a simple machining error could move the pivot point that far. That would explain why they shipped two identically bad lever caps in a row, and are having trouble shipping the fourth one.