Since you are transferring power, matched impedances always yield maximum output power. This always applies to linear systems.
Short answer
You have maximum power transfer when the voltage is 50% of no load voltage source.
So you are never trying to maximize voltage but achieve 50% of the no load source with a conjugate matched impedance.
Other misc ticky tacky info on MPPT
In the case of nonlinear source such as diodes or current sources, the maximum power transfer occurs when the incremental source impedance matches the incremental load impedance. For non reactive Z, we call this ESR and for nonlinear PV arrays, the MPPT usually at the cube root of 50% Voc or the Voc/Isc load line. but since Isc changes with solarity from 0 to some Imax the incremental source impedance has his effect of changing from the square root of 50%=70% for low sun power to the 4th power root of 50% = 84% for highest sun power. (from my research, sorry no ref's)
other
However source loss will also equal load loss, so minimizing both where possible improves on the maximum power for any given supply voltage.
You are converting DC power into mechanical piezo crystal motion and there will always be a conversion loss defined by the load crystal ESR (effective series resistance at f). Therefore minimizing cable and switch losses ensures you get the most output.
Piezo ultrasound devices like all crystals and ceramic resonators are defined by at least a R-L-Cs//Cp discrete parameters and not a simple resistor, so conjugate impedance matching is implied, ( which is a bigger topic) but the main concept here is to at least match the resistances. To examine your cct. the filter presents an impedance to the load of RLC of (10R+30nF)//1uH which resonates at 3MHz and |X(f)| for ~19 Ohms at 0 deg at LC resonance. The filter BW & rise time can be derived from Q = |X(f)|/R = 19/10 = 1.9 .. where X(f)= 2pi*3MHz*1uH = 19 ohms . Here a low Q gives wide bandwidth or fast rise time , a desirable characteristic.
We don't know Z(f)(piezo) or maximum power it can handle, so I cant judge what is optimal but sometimes, you don't want maximum power transfer but rather lower source impedance so that load variations do not affect the load voltage.
In this case the source is much lower than the load R you show, so there are other factors besides maximum power transfer such as load regulation just as in AC & DC power supplies where the source is always < 10% Z load and in the case of 1% regulation source Z is ~1% of load and in the case of the AC grid it may be 8~10% of the min. load Z but then network design with tap changers accommodate load fluctuations to give much better load regulation.
The reason matching is important is that power reflects back to the source if the interface impedances are not matched resulting in lower forward power. These are explained by an understanding of scattering parameters or s-parms such as s11 s21 and s22 the input return loss, transmission loss and output reflected return loss.
Having no load certainly gives you maximum output voltage but with no work being done and no power and energy being transferred.
Whenever there are time delays in a channel, reflections from mismatched impedances can cause image ghosting. But then your power efficiency is 50% at best. This how RF systems work in order to prevent echoes or ghosting or ringing in pulsed carriers. Since the load also depends on acoustic coupling of the ultrasonic waves to the target, and the target load impedance, this must also be considered when high resolution image artifacts or bogey targets must be avoided. This is one reason why fish finders cannot work close to the boat hull since that interface causes some strong reflections which must be masked. ( gated off)