Navigation information of airborne vehicles ejected from some moving platform need proper initialization and the process is termed as transfer alignment. The misalignment between the ejected vehicle and the moving platform is computationally estimated and corrected to provide accurate navigation. The estimator has to filter the noisy measurement provided by the navigational sensors mounted in the vehicle to arrive at the mismatch due to misalignment. Due to nonlinear system model, particle filter is designed for the estimation. However, presence of flexure effects in the daughter vehicle gets captured by the sensors and this further complicates the misalignment estimation. By using additional system states for flexure in the system model of the particle filter, the flexure can be estimated and corrected, but this increases the computational complexity. In this work, it is shown both analytically and through simulation that an evolutionary particle filter which evolves through multiple system models can provide at par performance in the steady state with lower computational complexity and better transient response.