The travel time distribution under interrupted flow based on radio frequency identification–detected data is analyzed. The urban road network studied is in downtown Nanjing, Jiangsu Province, in China, where video cameras and radio frequency equipment are installed at some arterial links to acquire traffic flow data including vehicle type, passing time, and spot speed. On the basis of the radio frequency identification data, the travel time distribution under interrupted flow shows a bimodal curve instead of a unimodal curve. This finding suggests that travel times under interrupted flow are not independent and identically distributed and can be divided into two components, the uninterrupted component and the interrupted component, according to whether vehicles experience intersection control delay. Six convex distribution models are presented, and an expectation–maximization algorithm is adopted to obtain the best distribution for each link unit. The results indicate that travel time under interrupted flow generally follows a bimodal distribution. The parameters of the models can distinguish travel times between those encountering delay and those without delay. This finding demonstrates that intersection control delay plays a predominant role in travel times under interrupted flow. Two main travel time reliability indexes under interrupted flow obtained by the parameters of the bimodal distribution, expected travel time and RATIO, are presented as applications. Expected travel times indicate the average levels of travel times of two independent components, and RATIO represents the ratio of the uncertain part of the travel times. The analysis indicates that these two indexes can capture the characteristics of interrupted-flow travel time reliability, which is mainly caused by link length, number of intersections, and mixed traffic flow.