There are two common ways to position an aerial survey precisely. Both use the same satellite observations; the difference is when the calculation happens, and what that timing costs you when conditions are poor.
Real-time: fast, and fragile in the wrong places
In the real-time method, corrections stream to the aircraft during the flight. When the radio link holds and the sky is open, the results are excellent and mobilisation is trivial: no marks to place, nothing to process afterwards.
When the link drops — behind a ridge, in a dead zone, below a rock face — those moments of the flight are simply lost. You cannot recover in the office what was never received in the air.
Post-processed: same data, better second chance
In the post-processed method, everything is recorded during the flight and the calculation happens afterwards on a workstation, run through the data both forwards and backwards in time.
That second pass is the whole point. A stretch of flight that could not be resolved going forwards can very often be resolved going backwards from a later fix. In a pit, a valley or a wooded corridor, post-processing recovers data that the real-time method simply loses.
Where we draw the line
Open ground, quick turnaround, everyday accuracy: real-time, delivered fast. Difficult sky, tight tolerances, or any dataset that will be argued about later: post-processed, with ground marks, checked against independently withheld points. It is a little slower and considerably more reliable.
The part neither one does
Both methods position the camera. Neither proves the ground. A survey with a perfect flight track but no independent checkpoints has no measured accuracy at all — only an expectation. Proving it is a separate, inexpensive step, and it is the one worth paying for.