Setting the pitch_moment_delta_elevator to -1.x solved the problem, but the
elevator became much more responsive. So apparently there is a bug with it and
also this parametre has a great effect in the modern FM.
I agree, either the doc is incomplete or there’s a bug. We’ll investigate. Did
you try tweaking elevator_effectiveness or elevator_lift_coef to ****
compensate this side effect?
After going through a somewhat similar experience with a very large airplane
with the reference datum being forward of the nose, I think the problem is (at
least with my experience) is that the description for the parameter
‘aero_center_lift’ does not state what it it’s value (the longitudinal
position of the aerodynamic center) is relative to. I had assumed it was
relative to the reference datum. That was a very bad assumption that led to my
airplane standing on its tail no matter what the CG position was in the fuel &
payload UI. It led to a huge offset that I could not understand where it was
coming from. It wasn’t until I displayed the sim forces that I could see the
forces were well aft of where the wings should have been. Since it is not
clear what the aero_center_lift parameter is relative to, it took trial-and-
error to get the wing forces positioned correctly, at least in terms of where
the forces are displayed relative to the wing model of the airplane. Still
need to check the debug screens to see if that is really correct. But at least
the airplane does not stand on its tail! Please include in the description of
the ‘aero_center_lift’ parameter what the value of this parameter is relative
to. It would also be nice for a description somewhere of how the value of this
variable affects the cg offset.
@Nocturne FYI
We went with a value of -0.22, which is as close as I could get to -0.02
without getting the reversal. Then we increased damping by 400 to compensate
for it. This seems to get a response close to what we wanted, but if this bug
is fixed, we’d happily revert back to what we had! 
Very interesting. Should this be something you would need to tweak? I mean, if
the Simulation was properly taking into consideration all the Wing Dimensions,
Sweep, Aspect Ratio, etc etc, it should position correctly the center of lift,
no? Never had to adjust those myself in any Aerodynamic Software I have used
before.
It may correctly position the center of lift if you have it compute it. But I
was using the recommended compute_center_of_lift = 0 for the new flight
model and specifying the starting position of the center of lift using the
'aero-center-lift` parameter.
My suggestion would be to retain it in the modern FM, as it really allows
fine-tuning the elevator’s response. Ideally, that bug outside the recommended
range should be fixed -it looks like a change of sign to me.