how is it possible for a model with maximum AUC to not also have minumum misclassification error?

Question

I have an elastic net model of a binary outcome where the lambda for max AUC is different than the lambda for min misclassification error. Shouldn't they be highly (inversely) correlated