Analyzing Canon ETTLii

These two images submitted on a NZ forum interest me [ ETTLii ]

That's a little crazy ! When you focused on the white object toward the front it exposed pretty well considering the subject was highly reflective .
But when you focused on the object behind it , a black object at a greater distance it under-exposed in comparison ?
The way I see it , one option is that for the background focus the camera did not register focus confirmation or it exposes any out of focus objects as average grey regardless of where they are and in this case the white object fooled the meter - strange ?

Update : I was just informed that the second shot was in fact manual focus and the camera did not confirm 'focus lock' which means we are back to square one - the point where we worked out that ETTLii drops the distance info and resorts to average gray metering when it doesn't confirm a focus lock- regardless of how far the subject is behind the white object .

Two more tests , this time auto-focus locked .... almost identical histograms due to the distance info being used .

I don't see any sense in Canon dropping the distance info when there is no focus-lock . I mean if a subject
is so badly out of focus why bother with changing the equation anyway ? If it is only out by a small percentage and could be a keeper when why not use the distance info anyway for a more accurate exposure ?