Maybe you run corners, parametric sweep or you have more interactive windows, and you accidentally annotated operating points and node voltages from 2 different run.
Otherwise it can be a bug in your cadence too, sometimes a simple reboot helps....sometimes.
On the attached screenshot is a nfet_rf transistor. I am pretty sure it model has an additional series S/D resistances. So, the dc OPs showing values for "pure" transistor only, while on the nodes at top level you can see sum of proper voltages and IR drop.
Interesting.... But why the Vgs for "pure" transistor is also different? It is 468.3mV, why higher than the annotated node voltage, which is only 460.6mV? Last should contain IR drop. I don`t understand....
Have you looked at the include-file where the models sit (which
may be a tangle, but proceed anyway) to see the structure below
the symbol (subcircuit-call) level? There is probably a source
resistor (and perhaps inductor, which may carry its own parasitic
R) accounting for the mismatch in values (the nodes are not the
same node).