Tonghui is claiming 120MHz!!
I tried that real quick with a VNA, just for the fun of it.
There are various options to hook up the test fixture to a VNA: grounding one end of the fixture and connecting the other end to the center conductor of a port (one-port shunt), or in series between two ports, or in a shunt configuration between two ports.
I opted for the one-port shunt configuration since it is the easiest to set up and it is best for impedances around 50 Ohms. So the right connector of the fixture was connected to a BNC short, and left one to the VNA port. Then full one-port OSM calibration was done in the fixture. As a match standard I used a 47 Ohms 0805 resistor. I did not go to the effort to tell the analyzer to correct for the discrepancy between 47 Ohms and 50 Ohms (open, short and match were assumed all assumed to be ideal), so that displayed values will be off by a few percent (impedances will read about about 6 percent low).
I tested the same 1.2 pF 0805 C0G cap as yesterday (capacitor-1p2.png). I got reasonable results although the whole matter is a bit touchy. You can see that the S_22 magnitude is above 0 dB at some frequencies. This is clearly a repeatability issue of the setup (slightly more losses when the cal standards were connected than during the measurement). On the other hand, we are at 0.005 dB/div here. I get a capacitance reading of about 1.2 pF up to 120 MHz. The LCR bridge measured about 1.3pF, but keep in mind that there is a systematic error involved (a quick calculation shows that this is roughly the expected discrepancy).
Next I tried a 1 nF 0805 X7R cap (capacitor-1n.png). This is less touchy than the 1.2 pF one, but again, the capacitance is reading systematically too low.
Finally I connected a TDK MPZ2012S102AT000 ferrite bead, intended for supply voltage filtering, in a 0805 package and with 1 kOhms impedance at 100 MHz (ferrite-bead.png). The impedance is just as expected, and it is nice and lossy.
As a sanity check I re-connected the 47 Ohms resistor used for calibration (load.png).
Depending on the DUT impedance a different connection of the test fixture might be better, and one could improve on calibration. But I would say they are not lying about the 120 MHz frequency range of that fixture when used and calibrated properly.
Edit: I repeated the calibration several times when I took these measurements. As you can see, the cal it a bit off in the measurement of the 1 nF cap since S_22 is almost at +0.2 dB at some frequencies. You really have to be careful about the calibration.