I believe that the human genome reference took DNA samples from different people and that some natural variation is included in extra contigs.
However, the human reference genome comes from a limited number of people.
Can the limited number of donors and the linear representation of the genome (as opposed to a graph representation that could potentially include variation) bias an analysis in favour of one ethnic group over another? Are there studies that have compared the alignment rates/mappabilities of reads of people coming from different geographic regions? Do some ethnic groups' reads align better to the reference?