Critical issues on coherent interface energy calculations revisited: The case of Al/TiB2