Urban areas are the most difficult case for a photogrammetry work. Usually a combination of oblique, zenital and detail pics should work but a good mission planning workflow is very important so you can optimise flights, pics and processing for a good result. I can't go into details but in short your usual planning software and clueless rectangular grid patterns won't work. Your flight pattern and camera angles should adap to urban grid, street width and building heights. I am also against doing it in a on single field data, you need to iterate in your planning approach, model photographing and software reconstruction loop. You need command of several software packages to do it right.
https://sketchfab.com/3d-models/montblanc-4998e976b9894acdaa93ade53075664fSome universities have published papers about city sampling and reconstruction works, but they use some serious overload of tools and resources we don't have, like combination of laser scanning on car with airplane based areal photogrammetry with rigs that get to shoot several cameras at once plus GCP with total stations, university support and staffing, etc. Even so, I think it can be done exclusively on photogrammetry work if you get to understand what it is going on, for instance how to get pics from ground level and from drone level correctly aligned.