This is expected generally, actually the DC sweep simulations do not include analysis of capacitors (They do include their parasitics if available in model.). But in transient simulation those capacitors are empty then in time they charge. Looking at DC solution you cannot say that the DC solution is the only possible solution for complicated models of MOSFETs. So what I suggest is that you extract the exact value of output voltage (As exact as your absolute tolerance allows.) then force your output to that voltage, then disconnect your output from that voltage see if it holds true.
But this measurement is kind of obsolote for an inverter as you wouldn't want to use inverter at that point. As an open loop system with high gain the 46 mV error is actually a good match between DC sweep and transient. Also a better way to look at the zero crossing point of an inverter is to connect its input to the output. If you do this and apply this voltage to the input of an exactly same inverter across a voltage controlled voltage source you can get better results.