If I give real measurements in the form of topocentric right ascension and declination from a ground station to an Orekit’s Orbit Determination process like the Batch Least Squares, I should NOT correct beforehand for the so-called aberration (a.k.a. light time correction, from the fact that the signal takes a non-zero duration from its reflection on the satellite to its arrival at the sensor, both moving) right?
I am asking as a look to the implementation of AngularRaDec seems to show that this effect is systematically included when computing theoretical measurements, with no way of disabling it.

Yes, you are right.
This is because the effect changes as the orbit is updated. When you want to perform precise orbit determination, you cannot rely on preprocessing removing this effect properly as it does not know the final orbit.

Thanks @Serrof and @luc for discussing this interesting topic. I was actually looking to correct for annual and diurnal aberrations in Orekit (as far as I understand they still need to be corrected for when retrieving topocentric RA-Dec from the astrometric solution of telescope images). I didn’t find how to do it in Orekit, and, while I know how to correct for the annual aberration using other libraries or formulas, I cannot find reliable sources in general on how to correct for the diurnal one. Is there some way to do this in Orekit?

Yes, you have to set up a custom implementation of EstimationModifier for the angular measurement.
We do not provide it yet in Orekit but it would be an interesting contribution.
This effect should remain separated from the raw AngularRaDec measurement and remain implemented as a modifier because angular measurements that are not based on relative direction with respect to background stars must not include this. So applying or not the modifier depends on the method used to perform the measurement.