Future space-based telescopes, such as the Wide-Field Infrared Survey Telescope (WFIRST), will observe the reflected-light spectra of directly imaged extrasolar planets. Interpretation of such data presents a number of novel challenges, including accounting for unknown planet radius and uncertain stellar illumination phase angle. Here we report on our continued development of Markov Chain Monte Carlo retrieval methods for addressing these issues in the interpretation of such data. Specifically we explore how the unknown planet radius and potentially poorly known observer-planet-star phase angle impacts retrievals of parameters of interest such as atmospheric methane abundance, cloud properties and surface gravity. As expected, the uncertainty in retrieved values is a strong function of signal-to-noise ratio (SNR) of the observed spectra, particularly for low metallicity atmospheres, which lack deep absorption signatures. Meaningful results may only be possible above certain SNR thresholds; for cases across a metallicity range of 1-50 times solar, we find that only an SNR of 20 systematically reproduces close to the correct methane abundance at all phase angles. However, even in cases where the phase angle is poorly known we find that the planet radius can be constrained to within a factor of two. We find that uncertainty in planet radius decreases at phase angles past quadrature, as the highly forward scattering nature of the atmosphere at these geometries limits the possible volume of phase space that relevant parameters can occupy. Finally, we present an estimation of possible improvement that can result from combining retrievals against observations at multiple phase angles.