It seems a pity that opportunistically mining an empty block is better than having a new block ready to begin, even before the contents and validity of the preceding block is confirmed.
In the latter case, it might be wrong, it might overlap, but at least some more TXs would be confirmed.