I think the default $PINS method works well for Calibre because it reduces the dependence upon pin order, which in turn reduces the need for information of pin order from primitive libraries.
v2lvs has the -i switch that you may be interested in. It results in a netlist that uses pin order instead of $PINS. In the Calibre Verification Users Manual there is a section titled "Using -i to Generate Simulation Output".
That may be what you need.