A question of flexibility in cytochrome c oxidase models