In fact, a recent paper by Menzie Chinn and Shang-Jin Wei presents empirical evidence that flexible exchange rate regimes do not aid current acount adjustment at all. If anything, adjustment in real exchange rates is slower under these regimes rather than faster. They write:
The assertion that a flexible exchange rate regime would facilitate current account adjustment is often repeated in policy circles. In this paper, we show that the assertion is not supported in the data.
Why doesn't the assertion hold? One possibility is that demand for currencies is a bit random. Chinn and a different coauthor argue that shifts in the volume of order flow track such random utility shocks. I think these shocks also track central bank interventions, as discussed in previous posts.