We conduct a comprehensive in-depth investigation of applied digital aerial photogrammetry to imagery-based inspection of large-scale photovoltaic installations. A varied collection of RGB and thermal datasets are collected and processed with different parameters, then analyzed based on multiple evaluation metrics. The analysis focuses on the image matching and camera parameters calibration phases, which constitute the backbone of the photogrammetric post-acquisition workflow ahead of defects detection. We emphasize the imbalance between RGB and thermal imagery, and how it leads to relatively very low computed and matched keypoints on thermal images. The keypoints t-distributed Stochastic Neighbor Embedding (t-SNE) from sample images highlights the repetitiveness of PV patterns, which increases the likelihood of image mismatching. The calibration of thermal datasets ends up with several inconsistencies in the optimization of the internal and external camera parameters. These inconsistencies produce flawed posterior results, which compromise the use of the photogrammetric approach. We develop state-of-art guidelines to adopt for image acquisition and processing, propose several research avenues to pursue for future work, and demonstrate the advantages that digital photogrammetry offers over conventional imagery-based inspection workflows.