Well clearly there is a difference. Also, obviously , No resistor is not yielding any results. I don't know why.
I think at this point you have three options.
1- your original setup which yields 0V for no load using only outside pins of sensor
2- 1 M ohm resistor using one outside pin and middle pin
3- 100 k resistor using one outside pin and middle pin.
I think you have enough data now to make an informed choice .
I would use either option -1 or 2.
The resolution seems to be better with option 1 because the range is double that of option-2
If you have a way to apply a known force less than max load (like in the midde) , I would compare the accuracy of options 1 & 2
using that test.