inductance of a piece of wire over a ground plane

I was doing some bond wire/ribbon analysis with HFSS. I could get slightly different results depending on the port configuration. Wire was bonded between two microstrip lines on 0.010" alumina.

1) I used wave ports and de-embedded out to the bond point. I analyzed again without the wire and then de-embedded these results in Microwave Office (microstrip open and pad capacitance) from the results with the wire. The final results should just be the wire over the substrate. Repeated for differnt lengths so I could get a plot of length vs. inductance. Got about 22 pH/mil. MWO bond wire model also matched results.

2) I used same wire shape but this time used lumped ports directly to the wire. The slope of the results (22 pH/mil) was exactly the same as case 1 however the extrapolated value for zero length was not zero as in case 1, but 0.2 nH, about 1 substrate height.

In case 1 (de-embedded out the microstrip and modal
ports) I am seeing the partial inductance of the wire and the
partial of the ground plane return (which should be almost
zero since it's a wide PEC sheet).

In case 2 I am seeing the full loop inductance (wire plus
the currents in the lumped ports. Extrapolating my results
back to zero gap show 0.2 nH when it should be 0 nH. The ports mujst have inductance (they do generate an H-field).

Also important to mesh inside the wire and also have a good mesh on the microstrip to achieve proper current distribution.

I concluded that for short wires (in this case a chip and wire matching circuit down to 1 ohm) that need high accuracy you have to be careful of the port selection.