There's a very interesting article somewhere (I can't remember where - I think Compass Points, but I couldn't find it yesterday reading the back archive), looking at sources of errors in mechanical instruments. Essentially they came down to offset errors such as the disk slipping, but also much more complex error caused by the jewelled pivot wearing out a (quite possibly non circular) hole in the card (particularly with the heavy clino - I seem to recall it mentioned that you have to do varying inclination readings when calibrating for this reason).
Personally I have not been able to detect any reproducible measurable difference (i.e. any calibration to apply) between the modern Suunto instruments held by Imperial, except for when the instrument is knackered to the point where it is retired anyway. As such I now just take multiple cross-checked measurements between them all to see if any of the compass cards are sticking etc.
However, we wear those poor little things out quickly taking them caving, so they're all pretty young - decades old club instruments may have been oddly magnetised while sitting in the library, balanced for the wrong hemisphere, used to bang in nails, badly jolted rattling in someone's car etc.
Graham - your WD compass certainly sounds like the effect of having some part of the enclosure becoming magnetised (I guess they didn't have such massive stocks of Alu or otherwise non-iron / ferromagnetic based metals back then?) - this happens inevitably to steel boats, and is calibrated by physically placing offsetting magnets in the binnacle.