Lifetime broadening is the quantum mechanical limit for bandwidth, but there are other so-called inhomogeneous broadening mechanisms at play in most situations, particularly for molecules. For example: unlike atoms, molecules occupy a large conformational space and every conformation can have a slightly different energy gap. A benzene can either be perfectly flat or ever so slightly ruffled. The slightly ruffled one will have slightly less conjugation and so a slightly shifted spectral signature. The ensemble measurement is a superposition of the individual spectra of each molecule in the sample that is probed by the measurement, and hence broadened compared to the spectra of individual molecules (which you can observe with certain techniques). Just so, atomic spectra are usually recorded at low pressure in gas phase, where the environment experienced by each atom is more or less identical and with no interactions with nearby neighbors. Molecular spectra are often recorded in condensed phase where there may be large variability in the local environment experienced by each molecule at the instant of measurement. The local environment exerts an effect on spectral transitions as well, which manifests as a broadening of spectroscopic peaks.
Note that even atoms are not usually natural-lifetime-limited. At any finite pressure there will be collisions and collisions reduce the excited state lifetime - hence collisions broaden the observed spectroscopic peaks. But, atomic spectra, especially recorded at low pressure, are a much closer representation of lifetime-broadened spectra than molecules. For this reason atomic spectra can often be fitted to Lorentzian or pseudo-Lorentzian (Voigt) line shape profiles, whereas molecular spectra usually have to be fitted to Gaussian functions, reflecting the inhomogeneous effects at play.