This photo shows what the patent refers to as the older Langevin transducer. It appears to be ceramic with two gold endcaps. The reddish wire is the drive wire. The photo tries to show an unusual characteristic of these filters. What appears to be a copper and green wire is actually just one wire which is green on one end and copper color on the other. This is done only on one end of the filter. The opposite transducer only has a copper colored ground wire. The patent never mentions why this might be done.
Also of interest is that the coils are glued around the resonator endcaps only to serve a strain relief function.

W9AC. Thank you for the paper. I was unaware that Collins filters looked like a 940 uH coil at the input. I suspect that is because Collins uses a magnetostrictive transducer. A variable capacitor was added externally to resonate with the coil forming a high pass filter which transforms 50 ohms to 145 kohms with the filter actually presenting an impedance of 12-16 kohms with resultant mismatch loss of 4-6 dB.
The Kokusai filter looks more like shunt capacitor across the piezolectric element which the patent mentions is “several kil-ohms or less” at resonance. I decided to create a lowpass matching network consisting of a shunt capacitor across the filter input which absorbs the transducers capacitance with a series inductor. With the ability to control both component values, I found that a 100 uH inductor with a 1000 pF capacitor could both resonate and match at the same time to about 1.7 kil-ohms. The result is below is almost a perfect match in the filter passband to 50 ohms. Unfortunately, this filter still has problems. It is more than a kHz too low in frequency and the passband ripple is severe. I suspect that the two problems are related but will need more investigation.

If you noticed the white strands attached to the weld for the ground wire connection to the adjacent resonator in the first photo; I tried your suggestion of a cotton ball instead of foam.