I Agee with 3v0, and you need to look at the system as a whole. The fact that you said a static load measures consistant tends to say it's not a measurement problem but rather might be a natural variation in the loading/stop loading cycle. When the measurement reaches setpoint and the valve is commanded to close there would seem to have to be some time lag that would allow some more small amount of material to continue to fall adding to the final weigth.
The fact that all your batch run values were higher then your setpoint, I tend to think that this is just going to be natural varitation in your batch process. You need to come up with a more complex control schem, such as having the valve start to close somewhat before reaching setpoint (couple of % say) and see how close to setpoint it can reach when material flow finally stops. There is bound to be some variation as things like material temp, humitity, density variation of material, etc, is sure to effect the final amount of material dropping once the flow valve closes.