WARNING: Extremely geeky electrical tech description 
What I saw was the open circuit voltage that the voltage divider was right at the threshold of what the microcontroller interprets as a “1” or high. Logic devices have a minimum input voltage that it interprets as a high, Vih-min, and maximum input voltage of what it interprets as a low, Vil-max. There is a “no man’s land” between the Vih and Vil that provides noise immunity by making sure input voltages are never in that area. Typically, that gap isn’t a problem as the logic chips are designed to work within those parameters. However, when connecting outside circuitry (the voltage divider, LED, etc.) the voltages presented are up to the circuit designer. I looked at the microcontroller’s logic signal specification (Vih Vil) and the voltages I measured were very very close to Vih. Unfortunately, these voltage specs are also temperature related so you can get intermittent behavior to further complicate diagnosis.
I chose the resistor value to give some margin above the voltage. It seems to be working. Removing the LED is probably a better solution in that it will totally remove that loading circuit. However, I found the LEDs very useful when I was building and wiring everything up. I tend to not modify boards in the off-chance that I have to replace the board and a formerly working system becomes a non-working system because I forgot about the mod (that’s just me, I have a short memory).
Hope this helps. Sorry if it got too technical.
s