I hopped on to Digikey and had a look to see if they had a datasheet for a variable cap matching the look of the one you have. I found
this one, and the
datasheet is here. It covers a whole range of caps, you'll have to scroll down to the page marked "10mm TOP/BOTTOM & SIDE ADJUST" on page 7 to see the 9-100pf part. As we suspected, the two opposing pins are the same, and both attach to the 'rotor' (the bit that spins), and the third pin is the 'stator' (the bit that doesn't).
It seems strange that it would behave differently in different ends of your circuit. I am by no means familiar with RF design, but it might be related to the Q of the capacitor. The 9-100pf cap from the digikey catalog uses a polycarbonate dielectric, and these seem to all have a Q of about 200 @ 1MHz. To put that in perspective, the Q of most smaller PTFE variable caps seems to fall around 1500 @ 1MHz, with some of the small SMT caps (see "7mm SURFACE MOUNT") sitting at 1000 @ 10MHz. It appears that the smaller package sizes and smaller ranges have a higher Q, and this makes sense given the small amounts of capacitance involved.
If you are delicately balancing impedance with and adjustable LC filter then you may want to use a smaller adjustable capacitor. This is all greek to me really, but I found a half decent reference for Q in radio circuits
here.
Can someone with more RF experience back me up on this?