My question concerns a statement by Godel in 1967. Commenting on Skolem’s failure to infer completeness from his (1922) proof of the Lowenheim-Skolem theorem, Godel observes that Skolem 



> did not give a correct proof of that completeness theorem which he explicitly stated (op. cit., p. 134), namely that there is a contradiction at some level n if there is an informal disproof of the formula. (Letter to Wang, in Wang 1974*, p. 10)<

 
Background: As I understand it, Godel’s own (1930) completeness proof proves the version: a formula A is either refutable or satisfiable. The proof relies on the method, taken from Lowenheim and Skolem, of “expanding” A into an indexed series of quantifier-free conjunctions of its instances, with existential variables replaced by functional terms (this is a gloss). These expansions are subject to the completeness of propositional logic: either some $A_n$ is refutable by the truth-table method, or $A_n$ is satisfiable for every n. In the second case, a model can be constructed from the satisfying assignments guaranteed for each $A_n$. In the first case, Godel shows that the satisfiability of A implies, for every n, the satisfiability of the expansion $A_n$. So, if $A_n$ is refutable for some n, then A is refutable.  

My question is as follows: in the quote above Godel states the theorem as: if ~A is provable, then there is an n such that $A_n$ is contradictory, "provable" taken in an informal sense. We can infer from $A_n$ being contradictory that A is not satisfiable (the quote continues: “evidently a correct informal disproof implies the nonexistence of a model”). If A is not satisfiable then ~A is valid. But this is a statement of *soundness* then, not completeness. Am I confused  about something?

*Wang, Hao. From mathematics to philosophy.  London: Routledge & Kegan Paul.