I’m not sure about that. I don’t think it does any math with “Always On” for “Unknown”, though I may be wrong. My understanding is Always On represents an average of the draw that was unattributed to devices within the last 24 hours. Unknown is the current draw of things that it thinks are devices (just unable to determine type). Since one is an average, and the other is the current draw, I don’t think they are related.
This furthers my questions about Always On. I really do wonder what happens if a device runs for more than 24 hours. If that was an identified device, does the power consumption of that device get reported twice (part of the “always on” from the previous 24 hours, and part of the named device). Either that happens, or Always On can some how differentiate and subtract out the “named devices”, but that doesn’t seem right either. Maybe Always On subtracts out the Unknown, and Unknown somehow subtracts out Always On, how does that idea work with this scenario? Maybe your initial hypothesis of “Under Reporting” is true, and I’d bet if you caught it right, there are times when it over reports (though I have yet to see someone bring that to the table with graphs supporting the notion).