Many crystallographic protein structures are being determined using molecular replacement (MR), a model-based phasing method that has become increasingly important with the steady growth of the PDB. While there are several highly automated software packages for MR, the methods for preparing optimal search models for MR are relatively unexplored. Recent advances in sequence-comparison methods allow the detection of more distantly related homologs and more accurate alignment of their sequences. It was investigated whether simple homology models (without modeling of unaligned regions) based on alignments from these improved methods are able to increase the potential of MR. 27 crystal structures were determined using a highly parallelized MR pipeline that facilitates all steps including homology detection, model preparation, MR searches, automated refinement and rebuilding. Several types of search models prepared with standard sequence-sequence alignment (BLAST) and more accurate profile-sequence and profile-profile methods (PSI-BLAST, FFAS) were compared in MR trials. The analysis shows that models based on more accurate alignments have a higher success rate in cases where the unknown structure and the search model share less than 35% sequence identity. It is concluded that by using different types of simple models based on accurate alignments, the success rate of MR can be significantly increased.