Even if basic multimeter gives you value in ohms, it does not mean that the value has anything to do with the soil you are measuring. I would say that negative values are even better, so its clear that you are doing something wrong.
Multimeter, in resistance mode, always generates current through the resistor you are measuring and does maths for the resulted current and voltage. If and when your measured thing, in this case soil, has any own electrical action going on, it will mess the result. With basically all multimeters, those currents both on the meter itself and soil are really close together, so that you would get better scientific results by throwing a dice.
Input impedance for voltage measurements has nothing to do with measuring resistances, two completely different measurements. High input impedance does “help” you to measure very low voltages, but as it does so, your measurements with different meters are not that directly comparable. Very small input impedance would short circuit your measured voltage and high one, like in most of the voltage measurement devices would let external errors in. As you have seen, just waving your test leads in air produces random voltages to your screen. And now, we are in situation that you might have 5 different voltage meters and all of them show different voltages, if you try them separately.
You could probably see the difference, if you set your highest input impedance meter in your experiment showing voltage from soil and then add your lowest impedance one directly in parallel with the first one. Voltage on first meters display will also go down when you add that second meter. If you now remove the high impedance meter from the setup, voltage on the second meter does not move up as much (if at all).
Sorry, this was me trying to be brief :D