In the context of signal detection in the presence of an unknown time-varying channel parameter, receivers based on the Expectation Propagation (EP) framework appear to be very promising. EP is a message-passing algorithm based on factor graphs with an inherent ability to combine prior knowledge of system variables with channel observations. This suggests that an effective estimation of random channel parameters can be achieved even with a very limited number of pilot symbols, thus increasing the payload efficiency. However, achieving satisfactory performance often requires ad-hoc adjustments in the way the probability distributions of latent variables - both data and channel parameters - are combined and projected. Here, we apply EP to a classical problem of coded transmission on a strong Wiener phase noise channel, employing soft-input soft-output decoding. We identify its limitations and propose new strategies which reach the performance benchmark while maintaining low complexity, with a primary focus on challenging scenarios where the state-of-the-art algorithms fail.