The values like -10dB and -18dB are not relevant criteria.
For it to have meaning, we would need to know the 0dB reference Sound Pressure Level (SPL) in dB.
The SPL at your listening volume should be measured instead, so for example you should be using a measured value at the listening position, such as 90dB SPL, and if you use a weighting, indicate which one (A, B, C, unweighted, etc).
For a standard reference you could also measure the SPL at 1 meter as well as at your listening position, and indicate the distance from the speaker you sit.
The output of the amplifier should be measured at the loudspeaker terminals without altering the volume setting with a voltmeter and converted to watts with a formula that takes into consideration the loudspeaker impedance. Since most voltmeters are unreliable above perhaps 100 Hz, the voltage measured should be considered as applying at that frequency and below, and will provide values usable for comparison with other systems and listeners.
It would also be helpful to indicate what kind of music you are using, since the crest factor varies with musical genres, and that will affect the average voltage output of the amplifier. Ideally we would all use the exact same recording for comparison, actually.
As the majority of power used by a system is at low frequencies, that is reasonably reliable as a measure of overall voltage output.
If you use the voltmeter to measure speaker system impedance, consider that the measurement will be at DC (0 Hz), not 100 Hz. You may be able to determine the impedance at <100Hz from a chart provided by the manufacturer, or a test report. If you do so, consider the lowest impedance in the chart that falls in the 1~100 Hz range as the impedance that uses maximum voltage output from the amplifier for your calculations.
2.83 Volts at 8 ohms = 1 watt output; at 4 ohms it equals 2 watts output, and so on.
If you go through the exercise, you may be surprised at how little power your amp is generating when listening at your normal volume level.
As for determining the peak power level on transients, that depends almost entirely on the source material. The voltmeter won't respond fast enough to read beyond an average level indication, and you will want to use a "true RMS" capable meter.
If you use a classical recording from an LP record, the difference between average and peak voltage could be as much as 20 dB, but if you use a modern pop recording from a digital source, it could be as low as 3dB and even 0dB ... the average level is the peak level.
