1. Figure 1 looks weird because the estimate of negative theta is smaller than true theta (and vice versa for positive theta). It is expected that it should be larger due to the use of normal distribution as a prior in EAP estimation.
2. I think the PR-SE method should be compared with contemporary ones like on-the-fly SH method, omega method proposed by Chen (2011) and others but not the old ones in this paper. The outperformance may only happen when compared with old methods but similar to newer methods. It is quite possible.