The efficacy of single distributions to model time headway data for mixed traffic with faster and slower vehicles is marginal. Effects of such traffic exaggerate on two-lane roads where both directions of traffic use a single carriageway. This paper aims to provide insights on modelling headways in this context based on field data. Shifted Erlang, and Exponential distributions can model headways under low flow when platoons are infrequent, and the proportion of shorter headways is insignificant. Moderate and heavy flow witness frequent interactions, platooning and significant slowing of vehicles. The use of mixture distribution exhibits aptness in describing the headways of the following and free vehicles; however, it entails a complex mathematical approach for parameter calibration, aggravating further if traffic is mixed. The paper introduces the concept of combined distribution as an alternative. It illustrates the effectiveness of the combined distribution function with shorter and longer headways modelled separately while describing headway data, respectively, at moderate and heavy flow. Lognormal and Gamma distribution models aptly described the shorter headways. Lognormal, Gamma and negative exponential models were found appropriate for longer ones. The combined models confirmed that the probability of shorter headways at moderate and heavy flow is relatively high compared to longer ones.