The key, I guess, is "the crudest level of description" - it's a descriptive view of grammar. Traditionally grammar had been a prescribed model - is my interpretation of your remark. But that's a false dichotomy, because prescription is in essence descriptive, but of a very narrow set of speakers (with the supposition to generalize it) - therefore it's abstract and restrictive while the master classical grammarian would still understand the grammar deemed incorrect, which recent theorists try to incorporate in the model instead.
The first quoted paragraph on glamour is somewhat confusing and does gloss over complicated developments of words that seem polysemous only because we write a bunch of related words all as one homophone but for ease of reading. For what it's worth, compare Ger. "Schreibe" - attributive noun of the verb "schreiben" ("to write") of the same discriminated *skr- root, or so they say, which also derives a score of other words sharded into different topics of which we can only scratch the surface which is scary complicated and prone to circular reasoning and visual language that is very graphic. In essence, taking in Dutch "begrip", Ger. "begriefen", "Begriff", "Grips", "Graben", "Gravur", and En. "gravitas", "grave" (ie. strength) should give an original sense like impart. Compare further Old English gren - "split" and Lt. "cerno" - "certain", "crino" - "think", "creat-" ... further "cran-" ... and "crane" compared to Ibis the totem animal of Thot in Egyptian mythology,, a diety related to wisdom writing and weighing judgement. Further compare OE dael, En. deal, tell, tole, Ger. *Zahl, Erzählen, to separate the wheat from the shaft. The -m in gram- needs explanation, still.
Traditional grammar tries a strict structure, but it takes different strokes for different folks, strongly strung together.