When the mtDNA profile of a crime scene matches that of a suspect, it is necessary to determine the probability of a chance match by consulting the frequencies of the identified allele in a "reference population". The ceiling principle suggests that that population should be chosen in which the allele of the suspect is found at the highest frequency, in order to give the suspect the maximum benefit of doubt. Recently, we advocated the use of a worldwide mitochondrial database combined with a geographical information system to identify the regions of the world with the highest frequencies of matching mtDNA types. Here, we demonstrate that the alternative approach of defining a ceiling reference population on the basis of continent or phenotype (race) is too coarse for a non-negligible percentage of mtDNA control region types.