Godel and use/mention

Wait a second: Godel Numbering enable us to map the formulae of a calculus onto the natural numbers. Isn't it then a type of semantics (where semantic interpretations enable us to map formulae onto other sets, most often {TRUE, FALSE} or {0,1} for example). Thus we have the Tarski problem of mixing our syntax with semantics, which is the Quine problem of mixing our use with mention etc.

I have seen people claiming this is flaw of the Godel proof but never followed it up. Maybe not a flaw but an illustration of the importance of keeping them separate!