I think we are wrong here and Oxford are right, my apologies.
I've researched the checklists for both the stable platform INS and the IRS systems and in each case there is a procedure for using the ATT mode after a failure of the power supply for one reason or another.
In the INS case you need to fly level for a period without accelerating to allow the gyros to align, the checklist says 'until the attitude failure flags on the ADIs are withdrawn'. For the IRS you fly level for 30 seconds, it doesn't say this but it would be to allow the IRS computer to detect the gravity vector, then you must input mag heading on the POS INIT page of the FMS or on the IRS control unit.
When operating in ATT both systems provide pitch and roll attitude and heading outputs.
Thunder Child, too complex and too paranoid! The JAA exams are generally much more straightforward than the old CAA exams. Most questions that appear to rely on dubious logic just look like that because of imprecise feedback or translation errors.
Concerning the 'shape of the earth' question it depends which nav systems you are talking about. The great circle and rhumb line questions and the concept of departure all assume a perfect sphere. Some older self-contained nav systems, particularly the Doppler based ones on V bombers, also assumed a perfect sphere. INS and Omega systems assume(d) an oblate spheroid whereas GPS and modern FMS nav systems assume the WGS84 spheroid. The statement you are talking about is usually made when you are teaching the very beginnings of navigation theory and is 'applicable until told otherwise'!