dB is a way to expresses power loss or power gain. For power loss, the formula is 10 x LOG (one power / another power both in Watts).
If the voltage is known for the same case the formula is 20 x LOG (one voltage, V1 / another voltage,V2) using the same resistance for the two voltages.
This is based on Power = E^2/R. The formula would be the same as 10 LOG (V1^2 / V2^2) where the R in the denominator being the same so it divides out (R/R=1).
The LOG is doubled from 10 to 20 because the power of 2 (ex V1^2) is dropped leaving just the ratio of V1 to V2. Same as saying 10 LOG(V1 / V2) + 10 LOG (V1/ V2) = 20 LOG ( V1 / V2).
Hope this makes sense. It is based on multiplying two numbers by finding the LOG of each and adding them together.
The Nanovna measures S11 using just Ch0 and this is expressed as S11 LOGMAG which is the return loss express in dB which equals 10 LOG (the reflected power / the forward power). The -35.19 dB you see is this amount.....a ratio of two powers expressed in dB.
The Nanovna also measures S21 (I think that is what you were referring to when you say S12). I think of it as the signal level at Ch1 (port 2) due to the signal applied from Ch0 (port 1)....in other words it is the gain or loss of something inserted between Ch0 and Ch1. In your case, -0.02 dB is possible at low frequencies for the cable you measured.
I think you did the calibration OK. For measuring the loss the "through" should be an adapter OR better yet the connectors chosen to be "insertable" or in other words one connector should be a male and the other female. The DUT should then have the appropriate connectors so no additional loss is added.
Hope all of this stuff I mentioned helps for the moment but I suggest doing some reading on S-parameters and logarithms.
