Aidan,
Indeed the code I am using does not include those lines. I will update my source code and try again.
As for the advice offered by Vikas, I think I am running it enough timesteps. I will however try it with a huge number of timesteps just in case. Regarding the sampling parameters is it generally better to have bigger correlation lengths (what the manual pages refer to as Nfreq)?
Needless to say how much I appreciate everyone’s help on this