Thanks all for the comments. I am not yet convinced that series inductance of the capacitor leads explains why I measure 26pF more at 28MHz. The 100pF was connected with minimal lead length so I wouldn't expect 60 to 80 nH stray inductance.
I agree that at low frequencies it measures the same as my LC meter.
I think Dave Eckhardt is correct and the answer to the original question is that the test fixture must be pretty rigid and stable even at 30MHz. At 300MHz you need something really good to accurately measure components. Ideally matched to 50 ohms all the way to the component.
Thanks Brian.
