Here are the BMARC repeater 2m cavities built in the 80s.

Cavities, the top cavity top fell apart, fixed (steel screws and brackets)

I thought the 1.5mm wire coupling loops were a bit skimpy.

1.5mm loops and inductor

So I used thicker wire and strip

3.2mm wire

But, it didn't work.

No low inductance (that wire) can get the pass and notch spacing better than 2MHz.
I believe (know) the thin wire has more inductance.
So, it looks like we need series inductance in the coupling loops.
(I did see that mentioned in the literature, suggesting a coil perpendicular
to the cavity's magnetic flux.)

Or, inductance from the loops to the N sockets. I'll try that.

added inductance

with inductance 1MHz cals.

One coupling coil in the cavity rotated 180, tried cap, poor, back to inductor

one coil 180, inductance

one coil 180, inductance, 1MHz cals.

Another consideration:
We use 1/4 wave coax cables for interconnect.
These give a further 3+db notch depth. So, we don't want stripline in the cavity assy., changing our 1/4 wave line lengths.

Cavity with the big black knob

cavity with knob, 1MHz cals. Thin 1.5mm wire, Philips beehive trimmer

The Wiltron 640

The Wiltron RF analyser