How much do smear-negative patients really contribute to tuberculosis transmissions? Re-examining an old question with new tools