What he means is (I'm sure he'll correct me if I interpret him wrongly) that whenever you build one building, you are naturally not building another. The time spent building the AO could be spent building saddlemaker (expanding kiln/whatever) and whatever benefit the AO might have on other buildings in the long run, your saddlemaker/kiln will still become productive later than it would be had you built it in the place (queue-wise) of the AO.
With a healthy dose of newb ignorance:
Personally I think the two proposed caveats somewhat cancel each other out, if it is advanced resource building you would be losing 'opportunity' time on, it must be considered that operating this advanced res building consumes resource, which being the concern of the first caveat reduces it's own opportunity value. If Res is indeed so short that you risk stalling the queue-at-large by virtue of prioritizing the AO, this will presumably also be the case with any AdvRes building.
The AO, unlike most tertiary buildings is always operational until such time as it will not be operational ever again (to all intents and purposes.) Thus the earlier it is built the more efficient, assuming caveat 1 is not true.
Regardless, it doesn't seem likely (to me) that one could conceivably lose more than one gains from an AO if built at a sane point in the schedule.