there are some well-known anomalies in scheduling, especially with a greedy algorithm, that require a dummy job (to avoid the strict greedy algorithm which is provably not the best in some cases) - see old papers of Ronald Graham (one is called bounds on multiprocessor scheduling anomalies)