Processing time is a function of a number of things - the number of images, proper overlaps, the plan you are on, the subject matter, how homogeneous the images are, elevation changes of the terrain, traffic on the server, etc.
You can think of photogrammetry as something like solving a puzzle. If the pieces are difficult to put together or vary widely, the engine will have to work harder to put the puzzle together. The more pieces, the longer it takes. Some subject matter like cities have clearer distinct reference features that are easier to stitch compared to fields or forests (more homogeneous - all looks the same) which take longer.
It is difficult to project how long a map will take. If it started processing on an easy part and we extrapolated based on that and then it got to a more difficult part where it slowed down, it wouldn’t be accurate.