I'm not sure how this chimes with "Do this force us to renounce to the idea of personal CEV [emphasis mine]? Hardly so."
There are infinitely many possible ways of extrapolating desires. But if you don't get around the part of "more the people we wished we were" (etc.), let's not call your musings on extrapolating CEV, because it doesn't fit the major criteria.
If an obscurantist (or anyone else for that matter) does not wish for his desires to change in any way, there is no personal CEV of him. Simple as that.
There may be other sensible ways of extrapolating / streamlining a utility function. It's an open question, and one that's much bigger than just CEV, the CEV part (as it's defined) is often answered easily enough.
Assume there's no personal CEV for certain obscurantists, then we are left with a theory that's supposed to tells us how to make people happy -- i.e. CEV -- and the example of an agent who cannot be made happy through their personal CEV -- i.e. an obscurantist; as the whole point of CEV is desire-satisfaction, if that fails to occur then the proposal isn't exactly fulfilling its role. You're correct that my musings aren't only on CEV, as they relate to the bigger question of what is a correct desire-satisfaction theory of well-being, which in turn might require figuring out how to extrapolate utility functions.
Consider the case of an obscurantist i.e. an irrational agent who is proudly ignorant and opposes the spread of (certain) knowledge even to themselves. First of all, is the existence of such an agent implausible? Not really, considering there are masochists out there and that, to some individuals, ignorance is bliss. How much, then, will be left of an obscurantist's identity upon coherently extrapolating their desires? The answers is probably not much, if anything at all.
Do this force us to renounce to the idea of personal CEV? Hardly so. Instead, do we decry the legitimacy of the obscurantist's desires? Perhaps, but a convincing argument must be provided for the ethical aspects of such a line of thought; a utilitarian could draw support from the societal benefits of increased epistemic hygiene in the absence of obscurantists.
In any case, this (admittedly contrived) example illustrates that there are pressing issues regarding CEV and personal identity. Also, on a related note, I recently heard a leading decision theorist say that their greatest concern with Ideal Advisor Theories was how desires become no longer the individuals' but, rather, those of their advisors; it may well be the case that personal CEV incurs in the same issues, at least under the obscurantist's conditions.
The limiting case above also reveals a subtle interplay between knowledge and volition; our desires might (implicitly) involve not wanting to know certain propositions, wanting to not know certain propositions, not wanting to act as if we knew certain propositions, wanting to act as if we did not know certain propositions.
What I just presented is not a rejection of the idea of personal CEV or similar desire-satisfaction theories of well-being, rather it aims to be a pointer to complications one must keep in mind when developing such proposals.