Ok so I spent all night experimenting and reading about this topic. Ashley was right in the end, although I think it could still be made easier for the end user by using a 0-1 scale.
From the Dr. Lex article, once again -
I sometimes get mails from people who want to know how they should configure a hardware or software volume control that already uses dB values by design. Some seem to believe they still need to apply a non-linear transformation to the dB values. No! The only things you need to determine there, are the range you want to use and the step size if applicable. For instance if the volume control offers a range of 120 dB, most likely you will want to limit it to the upper 60 dB range. Some controls offer attenuation (negative dB values) as well as amplification (positive dB values), you need to determine if your application needs either or both.
The key is to identify the lower bound - -50db in my case is more or less silent. This should be confirmed at the highest system volume available. If your lower bound is too low, then the lower parts of your volume settings will be imperceptible. My earlier (volumepercent*50)-50 is what I'll be going with moving forward.
Here's another article from Microsoft. When the online articles talk about linear controls, they are generally referring to linearly controlling the voltage or power - which results in the undesirable perceived volume, or loudness, curve as seen in the first diagram. It illustrates the problem of minimal effect on loudness at the upper volume levels, and disproportionally large effect at the lower volume levels. Where we're at, since we're working with decibels directly in Construct, is the second diagram already, where the loudness in db is linearly proportional to the volume control slider position. This is the desired behavior. Again, the key is identifying the minimum loudness, otherwise a portion of your lower volume settings will be useless.
So what happens if we apply a logarithmic function again to our decibel level in Construct? (the formula I've commonly consistantly found was 20(log10(v)). This would result in skewing the perceived loudness to volume position the OTHER way. That is, upper volume level adjustments would have a larger impact on loudness, but lower volume adjustments have a lower impact. While this is indeed skewed, it's not nearly as uncomfortable as the other way around since at louder levels we are able to hear more and discern more of a difference than when the loudness is very low. That's why some people have found using the logarithmic formula on the decibel setting acceptable or even preferable.
tldr:
1. Identifying the lower bound of attenuation is critical. I've found -50 to be a good result for my system at max volume, although lower system volumes result in differences in the lower end of the volume slider being hard to perceive, which is understandable.
2. Using linear decibal attenuation values results in linear perceived volume, which is "correct". Using logarithmic decibal attenuation values results in skewed perceived volume, which might be preferable, but "wrong" - this can be considered overcorrecting.
I've got an example project illustrating various approaches coming soon after I make it a bit more comprehensive and labelled.
TLDRTLDR:
Set your volume to -50*volumecontrol, where volumecontrol is a value from 0-1 and -50 is the point you don't hear anything anymore at max system volume. Mute when volumecontrol is 0.
PS: After all this I still think it would be nice for Construct to apply this for the user as a 0-1 value. Although the ability for each user to determine the lower bound themselves would be lost, a whole ton of confusion is resulting from users not understanding where that lower bound is to begin with, or how or why they should identify it themselves. Users who want to apply a further logarithmic adjustment can still do so on this 0-1 value. If the ability to apply more attenuation than -50 is desired to be preserved, then allow negative values. Values greater than 1 would be capped at 1 anyway, since we can't have less attenuation than no attenuation.