Due to the inherent autonomy and heterogeneity of Web services, ensuring composite services reliability remains a challenging problem. Extending the classical control flow with a transactional flow (encapsulating a set of recovery mechanisms) is widely accepted for ensuring composite services reliability. However, current approaches define recovery mechanisms in and ad-hoc way while they have to respect consistency rules regarding the control flow. In this paper, we propose a formal model based on event calculus for capturing both control and transactional flow of composite services. Following a pattern based modelling approach, we depict a set of rules characterising consistent transactional flow that can be defined w.r.t a given control flow. In addition, we propose two complementary approaches for enhancing composite services reliability.