The use of Remotely Piloted Aircraft Systems (RPAS) or drones allows for the generation of high-resolution photogrammetric products; however, the accuracy of the generated products can vary depending on georeferencing, the density of ground control points (GCPs), overlap, coverage, and topography. The objective of this research was to evaluate the accuracy of orthomosaics generated with RPAS using different georeferencing methods, numbers of GCPs, and overlap percentages in contrasting scenarios within La Libertad Metropolitan Park, Costa Rica. Eighteen photogrammetric flights were conducted in three scenarios-forest, open area, and highway-using a DJI Mavic 3 Multispectral (spatial resolution ≈2 cm/pixel). Three georeferencing methods were compared: integrated RTK, assisted RTK, and non-RTK positioning, with image overlaps ranging from 55% to 80% and varying numbers of GCPs. Flight imagery was processed using Pix4Dmapper, and accuracy was assessed based on the horizontal root mean square error (RMSH), vertical root mean square error (RMSZ), and georeferencing root mean square error (GRMS). RMSH values ranged from 0.02 to 0.58 m. The highest RMSZ and GRMS values were recorded in the forest environment, whereas the road environment exhibited the lowest GRMS values, ranging from 0.01 to 0.02 m. No significant differences were detected among the georeferencing methods, although assisted RTK improved vertical accuracy in the forest environment. The number of GCPs and image overlap significantly affected accuracy (p < 0.05); however, increasing these parameters did not result in proportional improvements. These findings suggest that optimized survey strategies could help reduce fieldwork time and processing costs without compromising orthomosaic quality.