Right. There is however an argument that the smaller total impulse a stage does (RD-58 is used for upper stages, where the stage mass is closer to the stage payload), the smaller the relative mass of fuel+engine of that stage is, and so the fixed mass of the bigger nozzle added to the stage costs more, in Tsiolkovsky equation, that the mass for the bigger stage. Going further, we can argue that for larger stages it's beneficial to spend some mass on nozzle, because that mass is used for larger impulse provided by the stage.

Interesting that using http://rocketworkbench.sourceforge.net/equil.phtml I couldn't get declared Isp (380) for methane-LOX for reasonable nozzle expansion ratios. Maybe the model is too approximate...