There are many environmental factors such as terrain, coastlines, night error (don't flight-test ADFs within 2 hours of sunset/sunup), storms, etc. However, if I remember correctly, at the maximum quadrantal error points (that's the error you mentioned which is worse on non-cardinal relative bearings), and within the promulgated range of the station, you should get a worse-case error of +/-5 degrees.
You should expect approximately a 10 degree error ("dip") in turns. That's just the way an ADF works.
As for 15 degree errors caused by other systems, which systems, specifically, are causing this sort of error ? (would it be RT on transmit or systems involve digital electronics, by any chance ?)
Would definitely recommend reading BCAR Section R (it's downloadable from the CAA site) - there's a section on testing ADFs. Even better - find a friendly neighbourhood licensed avionics engineer and ask them about it. They should know the required sign-off standards and have suitable ramp test gear for tracking down any tricky EMC problems far more quickly and cheaply than flight testing (though it's not as fun for the TP, obviously).
B.