You don't need to physically swap port 1 and port 2 to measure gain and solation. Just connect port1 to input and port2 to output.
RF2472G has a typical input P1dB=-10 dBm. This meas that if you VNA is generating -10dBm you will see a gain 1 dB lower than the actual (of course we are supposing P1dB of your device is actually -10 dBm).
Since the expected gain of the LNA is 14.6 dB, I suggest you to set the VNA to generate a power of about -20 dBm, that means an output power of your LNA of about -5.4 dBm high enough to have a precise measurement.
You can also set the VNA to generate at a fixed frequency a power sweep so that you can find the actual value of P1dB. Be careful not to exceed the absolute maximum ratings. For instance, you could set the maximum power to be generated to -5 dBm.
If you are interested to linearity you can do the measurement using two generators and a spectrum analyzer. In this way you can input to the device two tones having the same level at different frequencies (let say f1 and f2),. The you can measure the intermodulation product at a frequency of 2*f1-f2 then if the level of the two tones is Ptone and the intermodulation product if lower tahn this level of x dBc, then OIP3 = Pout + x/2.
If for instance you generate the tones with level of -20 dBm and you measure an intermodulation with power -50 dBm, then x = -20 - (-50) = 30 dBc then
OIP3 = -20 + 30/2 = -5 dBm. It is better to place isolators at the output of each generator in order to avoid that they intermodulates.