That is an tenable argument. We could however adopt another textual criticism device to explain the presence of the arupa pericopes. They were simply inserted inadvertently during the course of "redaction". Ven Analayo identifies (in his Comparative Study of the MN) quite a number of doctrinally ill-fitting pericopes, but shows how easily a reciter could have mistakenly inserted into a sutta, due to the preceding sequence of words.
I think more needs to be done to study this popular conception of "lateness" of a sutta. It's far too vague a concept right now to be of genuine utility. Eg, the insertion of the arupa pericopes only demonstrates that a text suffered an incorrect addition sometime "late", but it says nothing about the original text. Excise the arupa pericope, and we're still left with the issue - is the root-text "late"?
Another example is the oft-cited observation that texts with long compounds are "late". I would not disagree that compounding is a feature of late Pali, but it simply shows that single words were strung together at a late point in the text's history. It does not say anything about when these single words first came to be recorded, and it certainly is not proof that the concepts/connotations of a text were first set down in compounds instead of single words.
Yet another example is the fact that the Pali Canon shows heavy Sanskritisation, and this is taken as proof of lateness. Norman argues that the fact that the Commentaries can actually preserve the Prakrit roots and meanings of the Sanskritised Pali indicates that the textual tradition goes much further back in time than the Sanskrit period. Sanskritisation, is at best, evidence that the morphology of Pali closed late, but it says nothing about the age of the concepts.