There are two possible reasons why automated alignment procedures may fail to produce biologically correct alignments, (a) The chosen objective function may not be in accordance with biology, i.e., it may assign mathematically high scores to biologically wrong alignments. In this case, even efficient optimisation algorithms would lead to meaningless alignments. (b) The mathematically optimal alignment is biologically meaningful, but the employed heuristic optimisation procedure is not able to find the alignment with highest score. For the further development of alignment algorithms, it is crucial to find out which one of these reasons is to blame for mis-alignments produced by existing software programs. If (a) is often observed for an alignment program, efforts should be made to improve its underlying objective function. If (b) is the case, the biological quality of the output alignments can be improved by using a more efficient optimisation algorithm. For DIALIGN, it is unknown how close the produced alignments come to the numerically optimal alignment – in fact, it is possible to construct example sequences where DIALIGN's greedy heuristic produces alignments with arbitrarily low scores compared with the possible optimal alignment.