Little buckets don't work at all well, you need to be at least out into the far field, which for a large transducer at high frequency can be a LONG way (Think 10m or so), for smaller things that are less directional you need less distance, for some measurements (particularly things like beam patterns) temporal gating on the analyser is your friend.
Some companies offer test cells with absorptive foam lining that claim to emulate a large body of water, but I have no experience with these.
To get the transducers to wet out, spray with a dilute solution of washing up liquid before dropping it into the water.
RX off resonance is not a problem, just use a high impedance preamp close to the transducer, but I wish you luck getting much power into the thing more then about an octave off resonance, that is quite the trick (Note that many transducers have two usable resonances, sometimes with very different radiation patterns (Ring transducers with both hoop and thickness modes, looking at you). The acoustic matching between the ceramic and the water is often a 1/4 wave impedance match done with something like a carefully chosen thickness of FR4 in a simple device, or either a 1-3 piezo composite or something like a Tonpills element in low frequency devices.
Matching is matching, the standard RF methods apply, and usually you are trying to get a flatish response between the two usable resonances and maybe just a bit outside them.
There are two capacitors in the standard model of such things, the series one (Motional capacitance) and the parallel one at the input of the network (The 'Fixed' capacitance, which as its name suggests does not much change ,well apart from temperature and ageing).
They are actually not all that low current, you can have 10A circulating at resonance with a kV across the thing, hence the gapped core in the shunt inductor.
Regards, Dan.