Improving the preemptive bound for the single machine dynamic maximum lateness problem