The greedy selection of anchor points makes it possible for the user to prioritise potential anchor points according to arbitrary user-defined criteria. For example, one may use known gene boundaries in genomic sequences to define anchor points as we did in the Hox gene example described below. In addition, one may want to use automatically produced local alignments as anchor points to speed up the alignment procedure as outlined in [18]. Note that the set of gene boundaries will be necessarily consistent as long as the relative ordering among the genes is conserved. However, the automatically created anchor points may well be inconsistent with those 'biologically defined' anchors or inconsistent with each other. Since anchor points derived from expert knowledge should be more reliable than anchor points identified by some software program, it would make sense to first accept the known gene boundaries as anchors and then to use the automatically created local alignments, under the condition that they are consistent with the known gene boundaries. So in this case, one could use local alignment scores as scores for the automatically created anchor points, while one would assign arbitrarily defined higher scores to the biologically verified gene boundaries.