Standing back from the details, RSA does
$m^{ed}=m \mod pq$
Where e is a constant and d is the private key.
Now $e$ and $pq$ are coprime, so $e^{-1} \mod pq$ should be easy to determine using the extended euclidean algorithm. (Note, $\mod pq$, not $\mod (p-1)(q-1)$. Then trivially
$(m^e)^{e^{-1}} = m^{ee^{-1}} = m^1 \mod pq$
Somehow I doubt that I have broken RSA, so what is wrong with this analysis?
(RSA also uses multiplicative inverses in the full algorithm, but that is not my point.)