hfd - the original wording under JAR-FCL was that a test conducted by an examiner 'should' be observed. However, the CAA read it incorrectly as 'shall' be observed. I pointed this out at the pre-JAR roadshow; however Des P wouldn't accept he was wrong....
Since then, JAR-FCL has been amended and now also says 'shall' be observed.
Nevertheless, I think that it is OTT and we should challenge it when EAS part FCL NPAs are released.
I have no objection to the radnav fix during the diversion
once the initial visual estimate has been made. Being a bit of a softie (honest!), in the past I have directed a Skill Test candidate to do precisely that, knowing he was a few miles off his planned diversion track. He then fixed accurately and corrected to fly an excellent diversion after resetting the DI which was the initial cause of his error!