Not sure exactly what you're saying here, but that red word at least seems misleading to me. Surely you're not saying that the sum of both legs to ground is the same as from one lead to the other. Or else you're saying this is only with low impedance meters.
Or are you talking about when the inverter is operating, rather than when DC source is disconnected?
FWIW, the actual results I typically see when I test ungrounded (floated) arrays to ground with (I believe) a high impedance meter is a small voltage (<10 volts) that dissipates down to zero the longer I hold my meter to it. I've never really understood what's going on there, maybe you can explain. Sunpower modules show higher phantom voltages of this nature.
I am saying that if there are no leakage paths and the only possible current path from the DC wiring to ground is through the two meters, then the sum of the two DC readings to ground will be exactly the same as the line to line reading. Simple basic voltage divider.
If you add in additional leakage resistances, particularly unbalanced ones, then for a high impedance meter the sum of the two readings will still equal the line to line reading, even though the two line to ground readings may not be identical. If the meter impedance is low enough, swamping any leakage, then the two meter readings will be the same in magnitude.
Note carefully that I am NOT talking about two separate meter readings with the same meter moved from one side to the other. I am talking about using two identical (or different) meters connected simultaneously.
If you add in some hypothetical reference bias circuit that pumps the whole DC source up relative to ground by more than the source voltage itself, then you will have to be careful of the signs of the two voltages when adding them, but the result will still be the same.
(Depending on how you connect the meters, you might actually want to use the difference between the two signed readings rather than the sum, but the principle is still the same.)
The dissipating down to zero behavior you see with one meter implies that either there is a capacitance involved or a charge storage between the cell substrate and ground, of the kind that causes performance losses with some panels if the correct polarity of the string is not grounded.
I am not sure that type of panel can be used successfully in a floating DC system.