Choice of PCB coil design for filter inductor at ca 20 to 400 MHz, 5 W

Looking at the diplexer schematic from this nice question,

I realized this can completely be done in PCB-formed elements (minus the connectors). (As a kind of counter-action to the original question’s intent to get leaded capacitors, I’m not only trying to eliminate the leads, but also the discrete components.)

• the 50Ω aren’t actually resistors, they are the wave impedance of the cable / device you attach there
• a 20mm × 20mm rectangular plate capacitor formed by patches on the top and bottom layer of a 0.8mm thick piece of FR-4 ($varepsilon=4.7varepsilon_0$) is about 20 pF
• There’s PCB coil designs that provice up to µH in inductance.

Now, assuming I don’t plan to work with as high power as the original question, but still substantial power (5W) in a relevant range of 20 MHz – 500 MHz:

Would

1. a planar coil antenna
   
2. a coil formed through alternating between layers (so that the main magnetic flux is inside the substrate)
   

be most effective?

• I’m told (but I don’t have access to the software to simulate that) that planars have low Q. I have no such information on the other choice.
• Which design has lower self-capacitance?