Fixed bug: was reusing the same bounds after every metrization in reconstruct(), now restarting from the initial bounds for each metrization.
In metrize() final sampled matrix is now always reinitialising seed.