the wee turbine devices that are often used to measure the fuel flow are notoriously prone to miss-reading for the oddest reasons. I have seen issues caused by a simple bend in the fuel pipe leading up to the flow sensor, apparently the fuel flow in the pipe can become turbulent after a bend and this throws the sensor off. Also, if the electric fuel pump causes fluctuations or pulses in the fuel flow this can disturb the accuracy of the sensor readings. Another thing that we have found is that if the sensor is firmly fixed to something that vibrates the readings are thrown off and if it is mounted in an unsupported length of pipe it may vibrate at some rpms and this throws the accuracy off. In all of these cases the reading may be thrown off to a higher or a lower reading than in the optimum, smooth flow, straight vibration free situation.
If the reading is taken from the fuel metering part of a fuel injection system then all of the above is irrelevant.
Do the T's and P's change to suggest that the mixture goes richer when the fuel pump is switched off? It may be some other instrument error, perhaps the voltage in the panel changes when the load of the fuel pump is removed from the bus.
Rans6....