Your reasoning in favour of academies assumes that a lot of your commerce goes to science. Not always the case - I often push my economy to a point where I'm losing money at 100% gold for a time. Concentrating your science output in one city with settled Great Scientists makes sense here.
That is, why I said 1500 BC. i just looked up Mylenes latest normal game. Oracle gets built at 1700 BC, she has Alpha and Math by that time, so an Orracle -> Currency is possible, even on normal. (Btw,

come cheaper on Huge / Marathon as I heard and techs require more

, so one gets Techs later and Wonders are built earlier, so Currency through Oracle should be even more difficult on Marathon / Huge then it is on normal) .
With Currency, one can normally support the growing empire at at least 30% tech-rate (more with building wealth) even AFTER an ultra-expansion-phase of 1000y with science set to 0%. At least one should restart research there imho, to not fall to far backwards, and even with a Science-Slider as low as 30%, 10

(Capital Size 8 working Cottages + some TRs getting an Academy) is more then 7

if settling the GS. Only with a Science-Slider set below 20% for a longer time, settling would make sense, but running below 20% at that time is a big gamble imho. Believe me, I also expand hard, VERY hard, but loosing GPT at 0% Research? That's Overexpanding, and not good for more than 1000y and only if that are early 1000y where the AI also expands, if you do that later, you'll be so backwards that you won't be able to catch up / establish a tech-lead, though games can be won without that. Perhaps you're even more of a Warmonger then me.
Immediate vs. gradual rewards is a different beast and doesn't always depen on sheer efficiency. Returns on most things other than new cities (which put a lot of strain on your economy in the short term) are quite modest - consider 2 +2 cap raisers plus the food required to grow as the investment, merely to work 2 cottages or 1 mine supported by a farm. And this assumes you have grassland to spare.
Settled specialists and academies can afford to be a little less efficient than the others because 1) you have more control over where you get the bonus, meaning better multipliers 2) It's something you can invest above and beyond what's regularly available. We don't refuse to build universities just because they suck compared to libraries.
Where can't you chose to build an Academy and what does this have to do with the possibilty, that one can bulb a tech, worth 400-600 turns (Marathon, translates to 133-200 turns on Normal) in beakers of the settled specialist and have it instant? Bring an example, in which pure settling makes sense, and what does that argument with universities have to do with that? Universities are good, to build Oxford and then only if the game lasts as long as panzers or if they're speeded up by traits. I didn't even talk about the phase of the game, where Universities are available, that's past CS, I was talking about before that. Tell me, why don't you build Universities, besides that they're simply to expansive to build more than the required ones for Oxford.
Another consideration especially on faster speeds is that I don't want to speed up the global tech pace; absolute progress is less important than the time window for the gamewinning move. Diplomacy is also more awkward if you try to get ahead early, as opposed to optimising for the long run and let a sugar daddy/mommy carry my progress in the meantime.
Speeding up the global tech-pace comes through excessive tech-trade, what does that have to do with settling a GP, building an Academy or bulbing? As I hink, on faster speeds the biggest time window is with Cuirrassiers, as AI fails to tech them, bulb Philosophy with the 1st GS, build an Academy with the 2nd GS, run a GA with the GA from Music, get 2-3 more GSs, bulb hard in Education and Liberalism and you're done. Bulbing those, without having to trade for them, hightens the tech-pace for you while not hightening the one of others, so it's exactly what you want.
