There was a discussion a while ago about ththe PA design used
here
For a range of transformed impedances there are two configurations that you
can end up with. The reactance of the series L part of the
match can be absorbed into the tank coil, increasing its value, OR,
it can be absorbed into the series C DECREASING its value. Using this
latter solution, there will (usually) come a point with suitable choice of
tank Q where the reactance of the matching L exactly equals that of the tank
capacitor, making this disappear.
That is almost certainly what has happened in GW3UEP's design.
There are also a range of values where Xc tank is less than Xl match so it can
never be absorbed, and only the one solution is possible - that of adding it to
the tank inductor.
To do this you have to first get a working network with some arbitrary Q
chosen, and a valid output match for your optimum power output etc. Then adjust
the tank Q value while watching the resulting two possible values shown as C2 on
the spreadsheet. As Q is altered, one will move toward infinity then
disappear completely.
Andy G4JNT