The expected results are now more correct, but still not perfect.
This way, we can test more feature of them, in particular check for different errors, which were causing test failures.