Fine. Everything seems to work.
Now, for my example, we can say that "raw data" contains measured values ​​on the Earth's surface (ellipsoid), and result data displays the adjusted elements already on the projection plane.

There is only one caveat: the reduction is applied specifically for directions, but not for geodetic azimuths measured on the Earth's surface. To reduce the geodetic azimuths measured on the surface of the Earth, the program needs to "be able" to calculate and apply the convergence of the meridians. In the meantime, the program does not know how to do this, it is only possible to set such a value of the geodetic azimuths measured on the surface of the Earth by a banal sequential selection so that after reduction (as in directions) we get the required azimuth value already on the projection plane.

Sorry for my English )