It is in fact proposed, in Schane (1968) that French glides derive from vowels. Morin (1971: "Computer experiments in generative phonology: Low-level French phonology") pursues this further to handle some problems (formally restricting where the rule applies), followed by Dell (1972: "Une regle d‘effacement de i en français"). Lyche (1979 – "glides in French") criticises these analyses, but does not deny the existence of glide-vowel rules, instead she posits a different analysis which presages contemporary OT Output-Output constraints (alternativly, word-level cyclicity).
It is clear from these works and all subsequent discussion that glides in French are not allophones of high vowels, because of the "opacity" of the rules deriving glides. It is possible that French glides derive by rule from underlying vowels, but those rules are not allophonic.