First, as described, this is useless for both pubsub and MUC. It goes as far as describing an "item" subelement, but then fails to explain what this represents for either.
I agree the MUC example for item could be made more clear that the name in this case is the JID localpart. Or perhaps instead of saying "type specific" we could mandate that it is always a JID localpart. That would be fine for all my current implementations, but I wanted to leave room open for other types in the future to want name to maybe be something else. But honestly I'm not sure what else it could be so maybe saying it is always a localpart is fine.
One assumes it might be a room for MUC, and a node for pubsub, but what would then go inside each is entirely unspecified.
For MUC it gives a rather specific example of a <archive xmlns="urn:xmpp:pie:0#mam"/> child. Pubsub wouldn't use item tag at all, as it says it would use <pubsub xmlns="http://jabber.org/protocol/pubsub#owner"/> child just like existing PIE does for PEP.
1) Drop the "name" of the item, and replace it with "jid", "node", or both, to match disco#items. I would be inclined to specify that the first child is the payload (mandatory to understand), and any others are optional extensions.
I think calling it "jid" is confusing since it's just the localpart, and also component already has "jid" attribute. But I do think we can be more specific about what "name" means, per above.
2) Drop the type of the component in favour of "category" and "type", again from XEP-0030.
This is an interesting idea, and I think I quite like it. Though I'm concered that we don't actually have an identity category/type pair that uniquely identifies MUC (we have conference/text for example, but that could apply to any conference protocol).
3) Get at least the basics of MUC and/or Pubsub. I'd think MUC is config and affiliations, and Pubsub is config, affiliations, subscriptions, and items. I think all of those can be in the form they'd be returned by relevant IQ. Seems easy enough as a first cut, though undoubtedly implementation and discussion will surface more. All these get wrapped in a payload element to act as a schema support.
My current implementations hvae just archive and items, which is why that's all I specified in the document, but having eg the MUC config data form as another optional child in that case, etc, as you describe seems very sensible to me.