You want to make sure you perform a de-novo assembly. For bacteria, we have been using SPades with the error correction tool (using all kmers from 31 to 127). Then we use PROKKA to annotate the contigs (that is, to identify the open-reading frames). We get perfect profile matches (7 loci) over 99% of the time for Campylobacter and Listeria strains. Also, we have had great success with extended MLSTs.