A couple of observations based on your post here and Physics SE.
Your choice of antennas is a significant compromise. These are not balanced antennas so your coaxial cables will easily function as part of the antenna. This will cause stray RF and it will alter the receive and transmit patterns of the antennas. You would have more predictable results with a balanced antenna, such as a half wave, center fed dipole, and ensuring that your feedline length is an odd multiple of 1/4 of wavelength of 950 MHz. The directionality and impedance of the center fed, half wavelength dipole is well understood and documented.
The intent of your experiment is not completely clear. If you are trying to determine the effect of the sample on the attenuation or reflection of RF, you should position your antennas at least 1 wavelength (~0.318 meters) apart and insert your sample material mid position. You can calculate the amount of transmit power required by using a link budget calculator based on your receiver sensitivity, the antenna spacing, and the gain of the antennas. For dipole antennas, you can estimate the gain to be 2.15 dBi each. As a rough calculation, if your receiver has a -107 dBm (1 $\mu$V at 50 ohms) sensitivity, the minimum transmit power would be approximately -90 dBm. You can of course determine this experimentally by simply stepping down the transmit power until the receiver no longer registers a signal and then bring it back up just a bit.
Your test fixture should be in an anechoic chamber in order to eliminate unwanted reflections and standing waves. If you fail to take this step, you will likely find the results of your experiment to be inconsistent and unrepeatable.
The screen shot you showed of the VNA in the Physics SE link shows that you have the bandwidth of your VNA set much too wide. You should have it set to perhaps +/- 50 kHz of the carrier frequency instead of multiple harmonics as shown. This will allow you to more easily see the effects of the change of scattering parameters on the plot.