Well P2 and P3 are always going to show 180 degrees apart without the ground connection in between. That is probably not useful information.
I just now noticed that you are using a bowtie dipole, which is broader bandwidth and higher feed impedance than the wire dipole model that I used. I’m not sure how practical it is to use 1-port s1p data for half of a dipole, which depends highly on the common-mode reference to something, such as a reflector screen or just the shield of the feed line. Bowtie used to be fed with 300-ohm twin lead. I don’t know how you would measure one half of the dipole that way.
You might want to download free EZNEC and simulate the bowtie. You can probably find a close example design to get started. Get s1p for the balanced feed impedance, then use an ideal 1:1 transformer in uSimmics to feed it with a balanced circuit.