Indeed when measuring pulsed, distorted, ringing, harmonics rich signals RMS measurement will give you exact value you seek. In theory. In practice a lot depends on probes and their impedance as well as their bandwidth (also, do you compensate your probes from time to time- you need to do that), same goes for DSO but also higher sampling rate means more accurate calculation. Bigger buffer of DSO also means more accurate measurement over time (if you need one) etc. Differential configuration of two probes using math should help with noise to some extent - alway use that if possible.

The only other method that I know of that can deal with these kind of signals is by using RF thermal wattmeter. Relying on multimeters while dealing with these complex signals leads to inevitable false readings.

Also, you can easily measure output power on recovery coils using two probes and math function of your DSO (if your DSO has isolated inputs). So, connect the diode, capacitor, battery or whatever you're using to store recovered energy. Put shunt between diode and that "load". Connect second probe over that shunt (let's say it's CH2). Now, set appropriate values on your scope and use math function CH1xCH2. When you turn on your motor you will see power graph (on your DSO it's a red trace). Now, turn off CH1 and CH2 traces and turn on RMS measurement function. The RMS values of power graph will be presented in red letters. That's your recovered power in real time so when you change some parameter you will immediately see the effect on recovered power both in graphical as well as in numerical way. The same principle goes for measurement of input power if you don't want to use calculator- so connect one probe over the power supply (battery) and the other one over the shunt and repeat the procedure I described and you'll see power consumption graph. If your DSO have a hard time calculating stable RMS value simply set some higher value of timebase (to see more impulse signal on screen).

Now, if you had a 4 channels DSO with fully, galvanically isolated inputs you could do those two procedures at the same time and have both power graphs represented at the same time to see their relationship when changing some parameter.

However, if your two channel DSO has isolated inputs you could do the following- place one probe over the MOSFET Source shunt and put other probe over the recovery diode to "load" shunt. Use the same values for both channels and take both RMS measurements at the same time. You will have a relationship of two currents represented graphically as well as numerically so you can easily calculate input vs recovered power- RMS of course.

That's all for now, I'm still not feeling well and I really need to get some rest.