You're very welcome mate.
The weighted dB curves were created to measure "disturbing noise" or "sound pollution" so they're more useful to know if the noise from traffic, factories, railways, etc. would be heard and could disturb the people living around the source, than to determine harmful exposure. That's why they are calculated taking into account the different sensivity of human ear depending on frequency. Also this explains that when audio gear manufacturers offer their measurements of SNR they prefer doing it on dB(A). It's always more favourable to get high values, since frequently noise comes from PS hums that appear at 50-60 Hz depending on your country.
If I recall correctly, the A curve is more useful for sounds below 90 or 95dB and the C curve for sounds above that limit and it's more similar to a pure dB SPL curve. Other curves were used to take into account FM performance and Dolby filters in the cassette era. None of these measure the real pressure level, so disregard the true potential harm of the measured noise, specially at low frequencies, which is not negligible.
Accurate SPL sonometers are very expensive. They need very flat measuring microphones with a huge dynamic linearity. Just one of those mics can cost more than 2500 USD. They also need to be frequently re-calibrated.
I agree on your findings about the headphones sounding more effortless showing dynamic swings on well recorded (dynamically uncompressed) music. I think it's mainly for two factors, one of them sensivity, the other noise isolation. Using a transductor right in front your ear that is capable of 94 to 115 dB with just 1mW of power, is something that only very high sensivity loudspeakers (mostly horns) can match. With such sensivities any minute voltage change following the recorded signal can be converted into a change in sound.
OTOH isolation allows you to listen better the lowest sounds, so you don't need to increase the volume to hear them, and when a loud sound happens, there's dynamic headroom enough to preserve the relative scale. This is hard to match for most speakers. Using them even on silent environments, you wouldn't be able to listen at 10-20 dB what was recorded at that level, you more likely would be using 40dB. This means that when a sudden loud event happens, your 90dB headroom would require your speakers (and family and neighbours) to stand a 130dB sound. Very few speakers can play that loud, forget to do so without noticeable distortion. Add also the fact that having an average domestic loudspeaker (let's suppose 89dB/w/m) playing that loud, would require an amp capable to deliver more than 1000wpc, and that listening at 1m from the speakers
Compared to this, headphones can allow you to play 10dB sounds at 20dB so when that same dynamic swing happens, the loudest would be 110dB, which is well into their available range, reproduction is decently distortion free, and you're not asking huge power delivery to your amp, probably about 1mW, perhaps 64mW if using low sensivity cans.
Returning to the safe levels issue, I've found this graph for you at Wiki:
If your soundmeter allows C weighted measurements, use it since it's closer to the SPL curve. Just make your "measurement" CD. Create a frequency swept from 20Hz to 20KHz, let's say along 30 seconds and logarithmic if possible, at 0dBFS (the maximum intensity in a digital signal). Add another track of pink noise at 0dB FS too. No music would sound louder than that. Adjust the volume pot of your amp to have your meter reading the loudest you'd like to listen, let's say 100dB or 110dB, which is still safe for a less than one minute "fortissimo". That's the highest pot mark you should use with well recorded classical.
Repeat the reading and pot adjustment for 85-90dB. That's the level you should use for safe long listening sessions to "commercial" music, which we suppose heavily dinamically compressed.
You can repeat those settings and measurements for your different phones, so you'd know what volume setting to use safely with each of them.
Rgrds