There are a variety of ways you can process a dataset. Two basic points are that shorter baselines are easier to process and so is static data. So you really want to use the CORS only to transfer exact coordinates to your local base via static processing. Then you could via time editing process your rover file as static for each target. That’s the most robust way.
Now since your local base and rover are so close and the receiving environment seems to be very good the whole rover file could also be processed as kinematic with good results. I just retried that using only GPS and only in forward mode and got > 99% fixed. Assuming a correct fix any solution from the POS file during a static period will be within several cm of the overall average for that particular static period. Just another way of getting at the same thing.
In theory you could process your rover data with the more distant CORS as a base and get the same results, but that extra baseline distance can cause issues - fewer fixes, false fixes, in general more noise.
When I processed the rover data as kinematic I used no time edits in RTKPOST so I’m not sure why you are getting worse results there.
Basically to get really good at this takes some time. Practice helps and everybody has slightly different needs. Your proposed ideal scenario sounds perfectly fine. I’d say practice some more and if you have a specfic data set that you just can’t figure out what’s going on I can take a look at it. Also, look through the RTKLIBExplorer website if you haven’t yet as there’s a lot of good information there.
