"using VNA with a probe" - that doesn't make much sense, can you provide more detail or a photo of your test setup?
What was your test environment? You can easily get unexpected results by having your antenna too near a metal surface, or more likely not having the same physical ground structure at your test fixture as you have in your application.
Have you verified your test setup by measuring a known impedance?
A crude technique I've used in the past is to use two identical antennas, one as source and one as receptor and then make a couple of S21 measurements with the setup and extract the parameters from there. If you can get a decent S11 setup you'll probably get much better results though.
John D