When setting up a duplicate detection job, when run, the job by default is executed using all the Duplicate Detection Rules that are published. It would be great when creating a duplicate job that you can select which of the "rules" you want the job to run against - for example, detecting duplicate contacts, it makes no sense to run a duplicate detection of a contacts email against the Account entity. The current method provides so called duplicates because the contact works for the same organisation of another contact. Currently I get around this by "unpublish" rules and then execute the job. This is of course risky to perform during business hours.