seems like a long time for the CO detector to trigger - presumably the fire was turned off sometime earlier in the night. seems like a long delay to me? did they say what time the alarm went off, and if so, how long after the appliance being turned off did it go off? I don't suppose they've got any high level openings in the house have they (e.g. velux's etc)? if would suggest imho that if thermal inversion is the cause of the problem, an extractive terminal is not necessarily to make the situation safe under all foreseeable circumstances - they depend on the wind to help extract the pocs, while thermal inversion can occur even in relatively still conditions. I don't know what the job's like, but perhaps a possible scenario might be something like this? - fire producing some CO as result of combustion (as many ILFEs do - hence the nice solid fuel-esque flame picture (assuming of course, it's the flame picture intended by the manufacturer), fire goes off, people go to bed, flue cools, remnants of pocs linger in flue, people open window or velux upstairs, thermal inversion occurs, alarm goes off. it's a tenuous scenario of course, but maybe it could perhaps give another avenue to your investigation? a great deal of this will of course depend on the presence of any high level openings, and their relative postion to the flue terminal.