I'm persuaded that measuring total coverage isn't very useful. Coverage-per-test is much more interesting, but rarely implemented.