This can really be a tough call and that is why I use a judges panel to sort this out.
My own opinion (which has no bearing on the contest rules - they are already written and the judges, not I, will make the final determinations ..) is that if a fake language simply sounds like some identifiable CLASS of languages, plus its instrumental accompaniment is all composed so that you strongly think of something SPECIFIC, like "Wow, a traditional Faeroese sea shanty!" or "this singing plus that percussion REALLY reminds me of the beaches of Bali" --- BUT --- you the listener are not acutally processing sylables into possible meanings - THEN personally I'd still call it Instrumantal and that would be OK.
I had someone ask me about game music (Suikoden, I think?) and my research network could not pin it down, so I said OK. One or two people thought it was Tibetan, but they could not CONFIRM it, and I am pretty liberal in that regard - if there was no conlcusinve evidence, then "OK by me."
BUT: in the example above, the fake word "Archangibus" would (for me) actually start my grammar codecs working in Latin, and it would actually change the meaning of that moment - so I'd be more reluctant to call it 'instrumental.' The same thing happens with the fake prefix 'ur-' in Tolkein's made-up languages; it works exactly like the Latin prefix 'arch-' or the German prefix 'uber-' or 'ober-' and therefore it can create a meaning: such as geeks vs. ur-geeks. (I found about 506 hits for 'ur-geek,' btw. See how that works? It's turning into a 'real' word in English and in German.)
So, I guess I'd say, absent a specific example, that if individual syllables actually start connoting specific meanings, then it's no longer 'instrumental.'
But still a damn tough call especially with music like See-Saw!
- G!