I agree with Mazz, but let me add some concepts.
The standard used to calibrate the VNA sholu'be known for their "inperfections", for example a non-zero-lenght thru shoul'd be known for its delay, Zo, and Loss. This latter parameter is the most NON accurate parameter and impact strongly on the loss measurement of the low loss DUT (like connectors alone).
For example, for the popular 3.5mm , the most known factories that produce standards (Agilent, Anritsu, Maury) give no or very different informations about the loss.