Recently, advocates of DNA taxonomy have complained that there is inadequate control of the taxonomy in databases such as GenBank. This is correct, but the uncertainty may be extended to the sequences themselves. The present study shows that as long as vouchers are available neither problem is fatal, but if no voucher exists, bad sequences and bad taxonomy may be forever linked. Previous phylogenetic analyses of rbcL sequences have indicated that the small, south hemisphere, genus Oxychloë (Juncaceae) surprisingly either is embedded within or is a sister group to the Cyperaceae. This is not in accordance with traditional or current morphological data. By studying five new accessions, representing four species of Oxychloë, and re-examining the two vouchers of O. andina that were used in previous phylogenies, it has been possible to show that these two sequences are erroneous. One is a chimeric sequence and the other is most likely ''a contaminant.''