Ultimately, ethical imperatives must be absolute in nature. That is to say, it is invalid to use a relative measure to determine the morality of an action, unless that relative measure is used as an inference to some absolute state of affairs.
Suppose a person were to assert that "college students have a duty to get good grades". This imperative would be invalid if grades were rated on a curve such that some students must get bad grades. However, many people actually intend this statement merely as a surrogate for the statement that "college students should study hard and learn the materials to the best of their ability", with the assumption that some will inevitably not study hard and be the same students who do poorly.
Additionally, the assertion that "only the best alternative among good alternatives is morally acceptable" is nonsensical, as it represents a double evaluation. If an alternative can be seen as good, then it cannot follow that it is also not good. Typically, examples that fit this definition suffer from a lack of sufficient parsing, stringing together series of actions into a single object. It may actually be possible to define actions in such a way that better alternatives do not exist, by looking at each decision point as a separate action in which only one choice is the good one.