MSFS turbine modeling is pretty bad.
That accepted... the best we can do is piece things together and use what IS workable.
I'm no turbine expert, but I understand the differences between turbine and piston engines, when it comes to tourque and fuel-flow, and how they are effected by a constant-speed prop.
Your assumption:
I think that when reducing the RPM, the fuel flow should stay the same and the torque should change, not the other way around!
should be correct. Turbine fuel-flow is more throttle-position based, regardless of RPM; where piston fuel-flow
IS RPM based regardless of throttle position. If the mixture is proper, RPMs determine the amount of air that passes through a piston intake system, hence the amount of fuel.
You "throttle" a piston engine by controlling the air-flow. You "throttle" a turbine by controlling the fuel-flow.
Ideally.. torque is not RPM related. It's HORSEpower that is a function of tourque/RPM. However, the reality of piston/crank geometry makes torque a slave to RPM (ie. tourque curve). That's one of the main advantages to a constant-speed prop (especially on a piston engine). It allows you to stay on a specific part of the torque curve (RPMwise), no matter the actual torque (power (MP))
There is no appreciable geometry to a turbine.. it just spins..