sirs and madams,
if i may be allowed to add a small opinion / experience. i used to shun granaries all together but have learned its invaluable use through this forum through the likes of mr. speedy.
one sound strategy that keeps up with AI expansion is having a granary located at the core of the empire as this will allow the player to speedily establish the first one or two rings of cities. in these cities then, depending on terrain can be used to further pump out settlers (illustrating the exponential potential of a single granary) or to pump out military to make up for the void of military units that was caused during the time that the granary was built.
i would recommend a close look at the portion in mr. sp[eedy's site which shows how he micromanages a city to optimize settler production thus keeping up with the AI. and this is not puny feat considering the game was at diety.
lastly, at some point in the game, when expansion slows down, there tends to be hapinness problems with "granaried" cities. the options are to micromanage and find a combination that will stop growth all together at a pre-revolt stage (i.e. 4 happy 4 unhappy citizens) and adjust accordingly as you hook up more luxuries or build more citizen appeasing infrastructure.
as an added point, use the luxury slider only if it will have an empire-wide effect meaning it will allow allow a sufficient number of cities to grow past their current levels which would otherwise have gone into revolt without the slider. or to push a WLTKD. using the slider to prevent revolt in one city is not cost efficient use of money, it may be wiser to stop the growth or sacrifice a single citizen as an entertainer.
may i add also that in this sense, civilization mirrors real life. if you were a real ruler of a real nation, it would not be cost efficient to spend a certain amount on luxuries because one citizen demands so. such expenditure would be justified only if a majority were to benefit from it.
hope this helps.
squidd