Clustering only clusters the files locally. The next step is to select a clustered release and click on “lookup in the browser”. Find the correct release in the list of search results and click on the green tagger button. This will load the release in Picard and you can drag the cluster from the left panel to the release in the right panel.
Alternatively, you can let the scan do the work, which is a bit more automated, but that option has the tendency to scatter your files over multiple versions of the same release or even compilations.
Whatever option you use though, don’t make your batches too large. Dumping your entire library in Picard may seem to save a lot of time, but everything will grind to a halt.