This algorithm is correct because its search space can be all possible ordering of the tasks. Using the good search algorithm and skipping groups of schedules that are known from bounding information to be worse than previously found schedules we rarely have to search all orderings. We know that the lower bound information is not too high because it uses the smallest values from the remaining tasks and fits them on the processors in the best possible way, no possible ordering of the remaining tasks could be better.