Hi Hubert,
I think your work flow is good. There are only a few things to consider.
If you use multiple stations, each station will have it own (small) frequency offset. These may need to be solved first. This is best done fitting all parameters of the TLE to one single station at a time and note the frequency. Correct the frequency of the points of this station to the nominal frequency. Do this for all stations. This allows to calibrate all the stations to each other. Next use the calibrated points to generate the final TLE. I hope that this aditional step makes it more robust.
Also try to fit over a longer period then 24 hours, especially if you want to fit inclination or drag value. For inspiration read @cgbsat his posts in this thread: Observation 1037954: Raw data Needed - #8 by cgbsat
Also the tabulation helper is good, but as you are working on a TLE corrected waterfall it is not the most accurate. Try RFFFT directly on your system, or setup a second SDR with an omni antenna, to run RFFFT and RFPLOT.
Hope this helps! best regards,
Eelke.