Have you tried measuring the continuity across each pair of contacts in turn? in theory you will be getting a high reading across each contact if they are indeed faulty. If they both read out negligible <0.01ohm then it will point toward your Zs testing method.
Is it possible that the no-trip method of measuring Zs with your instrument is picking up some inductance-related discrepancy?
The principle of an RCD is two coils on a ferrite ring, one in each pole, balancing a magnetic field and a sense-coil picking up any imbalance. It may be that different manufacturing methods of RCDs cause problems of varying degree with different makes of testers.
And keep in mind that if you test at the incomers to the RCD using no-trip, the RCD is effectively out of circuit so whatever your tester normally does to "quieten" the RCD may now be interfering with the result. Just a thought...