srnet:
And going back to the question posed in post #1 ?
Apologies, I must need more sleep.
Solved! ...sort of...
Welp, on my drive to work today I had a thought. I've been measuring the current of the entire circuit...including the boost converter. (I was originally testing total current consumption to establish battery life when I discovered this oddity)
I bypassed the boost converter by applying 3.3V from my bench supply to the pad with the boost converter output capacitor and found very different current measurements.
LED Off = 1.1 mA
LED On = 9.6 mA
Exactly as it should be.
It seems that my boost converter is horribly inefficient and draws ~30 mA to light a 8.5 mA LED. 
Seems so obvious now. I was stuck on the idea that this was happening at the ATMega328P. Guess it's back to the drawing board on the boost converter design.
Thanks so much for the help! I learned a few things along the way.
Buck converters tend to be significantly more efficient.
Thank you for the follow-up.