Hi Drew,
1) Where did I get the design from?

Magnetic Coupling
===========
http://www.unixservice.com.au/hamradio/repeaters/duplexers/IMG_3242.JPG

My pixture, inspired by the same in Motorola documentation on their T1500 UHF cans

This comes from the Fundamentals. The original 2M cans from the 60's used two coupling
loops and, couple some energy with an L or C between the connectors.
Note my word "energy". Cannot the energy be transferred Magnetically?

Next is judging the correct coil sizes.

Actually, that's not that hard. The Main loop area is the same as in a "normal" coupling loop.
The two loops were placed close together but the coupling was not adequate so I tried the coil
as pictured and, instant success.

2) How are the wires connected? Just two loops, Two, GND to connector pin, that's it.

3) Resonance...., There is NO resonance. With this design, there's only one resonance and that's
     determined by the cavity.

4) Adjustment. Yes, bend the wire to get close then twist the whole assembly in the in situ as is
    normal for this type of probe assembly. It's actually quite easy.

With this design we only get one Pass and one Notch, thus the notch is at least 3db better and
in practice as seen here, 10db better. Also the out of band, say 5MHz up and down, attenuation is
much better than the series L and C to GND or your two turn coax (hard-line) and C or the parallel
resonant L & C between two connectors.

Back to the question, where did I get the design from? Answer: Thinking laterally, getting out the soldering iron and
'aving a go'. The value of having time and no deadline. Seeing the value in previous work, realising I've
test-gear those in the 70s never dreamed of. (PocketVNA, Wiltron 640) Then comes the stroke of genius in swapping
the pass and notch frequencies, that did take a few days.

Now, I want this knowledge out in the Public Domain so it can't be patented.