Distortionless superdirective beamforming is widely used in reverberant and diffuse noise environments due to its high directivity and robustness in extracting the desired signal. While traditional metrics such as the directivity factor (DF) and white noise gain (WNG) are sufficient to characterize the performance of fixed beamformers, they do not account for the perceptual quality of the enhanced signal. In scenarios where objective evaluation of speech quality is needed, the choice of the target signal in beamforming remains unclear. In this work, we address this issue by examining three types of target signals: the direct-path component, the direct-path combined with early reflections, and the direct-path plus early reflections filtered by a reference beamformer. We take WNG-constrained robust superdirective beamformers as an example to analyze the impact of each target signal on performance. Our results show that using the direct-path combined with early reflections as target signal does not reliably reflect the beamformer’s ability to improve speech quality. In contrast, both the direct-path alone and the version incorporating early reflections filtered by a reference beamformer offer more accurate reflections of beamforming effectiveness, yet still fall short in reliably evaluating overall enhancement performance. These findings underscore the need for a more appropriate evaluation framework for directional filtering techniques.