But what you lukin at are lukin, sights, visions, visible things, so the 'moku e moku' rule still holds. They are, to be sure, often more than just that, but these further things are not relevant to their being seen. I think the corporeality of things seen gets in the way of noticing that, as seen, they are like sounds ('kute') and textures ('pilin').
The issue of where to put the references to the language seems to ride a lot on whether we translate 'toki' as "say" or "speak" (at least the arguments seem to go that way). Since the basic meaning is "communicate" and that goes with the 'kepeken' format, that line of chat seems settled (though the adverbial pattern is still possible, if not used for topics).
The "about" preposition (or whatever) still needs consideration, but, in the meantime, the 'ijo [topic]' form works, not only with 'toki', but also 'sitelen', 'kute', 'pilin' and 'sona' (and maybe some others I've forgotten). The adverbial approach is still also possible, I suppose.
As far as I can trace, the history of the 'nanpa' rule seems to be that it was officially 'pi nanpa N' in Pije and early corrections were made to that rule. But these corrections were regularly ignored. Eventually Miss Gadgrind bowed to Norma Loquendi and the corrections stopped being made. Then the parallel with multi-word cardinals came alone as an ex post facto justification for the change, and the corrections -- always, I think, in the form "you don't need 'pi' -- switched sides.
On 'nanpa', the question arises whether "how many?" should be 'nanpa seme' or 'mute seme' (both needing 'pi', by the way). Both seem OK and no difference of meaning has emerged. But I wonder if 'nanpa seme' without 'pi' might mean "which one in order?"