Well logic suggests that the cylinder head bolts have NOT been torqued the same for some considerable time, so whilst yes you are correct, clearly in practice SOME tolerance is permissible as it's been running generally fine albeit with a minor leak.
Presumably this is what's been causing the leak.
I took the approach that undoing everything and redoing it all from scratch wasn't something I wanted to do as I could make the problem worse.
It was obvious that just one nut was looser.
So logically I thought that if I could ensure that I didn't go over the torque on the good nuts then getting close to them would be better than it is now.
So I set about trying a low torque setting on the good nuts to get a feel for where they were. E.g. a setting of 20NM didn't budge the good nuts but turned the loose one. Then moved upwards from there without actually moving any of the good nuts. I got up to about 40NM.
Again the situation before was such that one of the nuts was less than 20NM so I can't have made it any worse.