There was probably enough stray capacitence from the unconnected neutral to ground and the
three phases. The three phases would cancel out to near zero volts. If the impedance of the DVMs is 10 megohms, and the stray capacitence is around 0.002 microfarads the DVM would measure about 99% of the true voltage. The voltage across a cap in series with a resistor (the DVM looks like a high value resistance) would be 90 degrees phase shifted compared with the voltage across the resistor. The voltage across the capacitence is around 15 volts. The 90 degree phase shift is why 15v + 117v not= 118v.
High impedance digital meters are great for electronics work but for what you're doing you probably want lower impedance. You could parallel a 100 kilohm 5 watt resistor with the DVM to get the impedance down. Oh, you'll still see some voltage, but the numbers will be so way off (like 50V) that you'll know something is broken.
