Generating a JPL SPICE kernel using orekit ephemeris

Hello everybody,

After some studies, I have corrected some bugs and simplified the code to generate an asteroid ephemeris.
Now JPL and MPC produce almost the same results. I guess the differences are related to the orbital perturbations considered and the different time scale (JTDB in JPL and TT in MPC).
The configuration used in JPL and MPC are in the attached images.

My code using Orekit almost generates similar results in X and Y axis but not in Z axis. I have tried different time scales (see below): TDB, TT and UTC with almost the same results.
The problem with the Z axis is not new and was originally discussed at.

a) Asteroid propagation not matching MPC results
b) Ecliptic frame, J2000 sun-centric - #3 by evan.ward

In fact, I’m using the framework defined in the last post, but the Z axis still have a different value.
I guess now the problem is in the frame definition. This is the information I have about the frame definition in JPL, MPC and my code.

Any help/clue would be appreciated.