The osteological paradox: problems of inferring health from skeletal samples