Coplanar strip lines length results discrepancy

Hello,

I am designing a simple asymmetric coplanar strip lines (widths are different for each strip) using Ansoft HFSS v12.

Dimensions and specifications:
Length of strip lines: 20mm
Width of one strip: 2.835mm
Width of other strip: 2mm
Thickness: 0.017mm

Substrate relative permittivity: 2.2 (Rogers TLY-5)
Substrate Thickness: 1.575mm

Radiation Boundary: air

I am using two waveports on either ends of the strips, normalized to 50 ohm.

Goal: to find Z0, characteristic impedances, of strips at the above values.

I am attaching the formula used for theoretical calculations. Theoretically putting the above values in the equation should result in 138 ohm. I have used Matlab to get this results. The formula involves ellipitical integrals.

But when simulate using HFSS v12 and plotting Z parameter (Z21 - real and imaginary), I am getting a spike for the real part, spiking upto 3750 ohm at 6 Ghz
and the imaginary curve spiking up and then down, cutting the zero-line axis at
nearly 6 Ghz. Please see the attached results for full understanding of the results, since picture is a thousand words.

I am also attaching the simulation file so that you can see it for yourself, possibly I dont think I am doing anything wrong or missing anything. If I am, please do let me know.

I do not understand this discrepancy between the HFSS results and theoretical ones from matlab. Any help or explanation will be appreciated.
Thanks in advance.

I had met a spike of HFSS simulation in another project because I ignored the grounded. When I grounded correctly, the spike disappeared.