Sounds like a common situation where L-N loads on "One Side" of the Xformer's coil have very little Capacitive Reactance [Xc] - loads are more true power, and the other side has high Xc loads.
The higher Xc side would have higher voltage, and the most likely culprite would be Induction Motors running with very light loads [or no loads] on their shafts.
Another reason would be a Transformer with high Impedance [like + 5.0% Z] which has a poor load balance deal going on.
Anyhow, this should not be a big problem until the RMS voltage under a load situation begins to exceed 135 VAC L-N and 260 VAC L-L [using a high loading effect volt meter - one with a low input Impedance].
That's when the smoke begins to spill out Always remember those DVMs might show above RMS values, with a response to the level of current drawn at a high input Impedance - equaling out to a higher than actual voltage readout!
Please pardon zee bad spel-eeng!