This paper considers the network- and application-level reliabilities for connected vehicles under path loss environments. We derive a probabilistic framework for estimating reliabilities that is applicable with various path loss models. We also build a realistic connected vehicle traffic simulation environment and use it to perform extensive experiments considering semi-urban traffic. The results show that the achievable reliability performances differ significantly with the path loss models considered. For a moderate communication distance between a transmitter and a receiver, with established deterministic and stochastic path loss models, the network-level reliability is around 55%, whereas with the realistic path loss models that consider obstacles and traffic, the reliability falls below 30%. To improve the network- and application-level reliabilities, we propose a feedbackless relaying mechanism that can be deployed on top of IEEE 802.11p, where, the relay vehicle selection is done autonomously. Here, the relaying mechanism improves the network-level and application-level reliabilities by at least 35% for the studied path loss models.