Okay, so count up how many times he’s given an infinitive-like {-meH} verb describing a noun vs. how many times he’s given a verb with {-meH} describing a noun a subject and/or object. It’s certainly a strong trend, if not a law.

Generally, looking at the use of {-meH} in canon, Okrand often makes a full purpose clause to describe the purpose of the action of a verb, but he nearly always uses a lone verb with {-meH} to give the purpose of a noun.

Do we act prescriptively or descriptively? Do we take a single counter-example as cause to reverse an observed trend and forget it ever happened?

Don't throw the prescriptivist argument at me, mister "I know what Okrand said but it's just ugly." Okrand nearly always uses -meH to modify verbs (sentences) instead of nouns, so our pool of unambiguous examples is too small to declare a trend. But he has used it to modify nouns, and in both places he has used it with and without arguments.

What we do is not tell people that they've done something ungrammatical when they clearly haven't. If you want to always use "infinite" purpose clauses on nouns and never "split" genitive phrases with modifying clauses, that's your prerogative, but there's ample evidence out there that the rules you're claiming aren't actually rules; they're just patterns you're overgeneralizing.

-- 
SuStel
http://trimboli.name