The use of directional room impulse responses (DRIR) measured with spherical microphone arrays (SMA) has become widespread in the reproduction of spatial reverberation effects on surround-sound systems through multichannel convolution. However, the measurement of such DRIRs in "real-world" conditions is inevitably subject to several risk factors, including the presence of a nondecaying noise floor that can produce an "infinite reverberation" effect when convolved with an input sound. Recent work has focused on model-based techniques for removing this noise floor by replacing it with a re-synthesized prolongation of the measured reverberation tail, which has concurrently led to the development of a framework for the spatial analysis of reverberation properties. We present here a comprehensive evaluation of the proposed techniques through their application to DRIRs measured in particularly complex spaces, including spatially anisotropic late reverberation tails as well as multipleslope decays characteristic of coupled-volume configurations. Following a brief review of the theoretical underpinnings of the reverberation tail re-synthesis denoising procedure, the measurement, treatment, analysis, and subsequent denoising of these DRIRs are each detailed and assessed with quantitative metrics. Finally, a discussion of the anisotropic, direction-dependent analysis results obtained is included as the basis for a wider research question on the acoustical considerations behind a stochastic reverberation model allowing for spatial variations.