There's a category error here; an elaborated-type-specifier is not a declaration.

[I think the existing wording was attempting to refer to this portion of [class.name]p2: "A declaration consisting of class-key identifier; [...] introduces the class name into the current scope." but generalizing this to apply to other cases in which elaborated-type-specifiers introduce class-names seems like a good idea.]

To some extent the problem is that this whole paragraph is bogus: "A declaration is a definition unless" is one big category error. A declaration declares zero or more entities, defines zero or more entities, and introduces zero or more names. But it's wrong to say that a declaration "is" or "is not" a definition.

Perhaps a different structure would help. For instance:

Each entity declared by a declaration is also defined by that declaration unless:

• [...]
• the entity is introduced by an elaborated-type-specifier,
• [...]
  A declaration is said to be a definition of each entity that it defines.

Or perhaps we need to formally divorce the notion of "declaration" (the grammar element) and "declaration" (the portion of a declaration that declares a specific entity).

I like the "different structure" approach, absent the major task of doing the divorce.