First off, I want to compliment the process you used to arrive at the final solution. Good photojournalism.
Something is definitely amiss with your MC-462A though. With an ambient of ~25C, the resultant CPU temp is way out of whack. Do you absolutely have to use that shim? General wisdom has it that they cause a slightly higher temperature. If it interferes with a tigh mate between the MC-462 base and the core, that would explain the extreme temperature. You do not have to worry about the MC-462 damaging your CPU. Do try again without the copper shim. You can increase the accuracy of the KT7A in-socket thermistor substantially by bending it slightly upward so that when the CPU is in the socket, the head of the thermistor rests firmly up against the bottom of the CPU. Before you lock the zif socket the CPU should ride slightly higher because it rests upon the lifted thermistor head. You push the CPU down into the socket and then lock the zif lever. I put a small dib of AS on the in-socket thermistor head, right where it contacts the bottom of the CPU, for even better thermal conductivity. Using that approach, the in-socket thermistor reads within .5C of the on-die thermistor once the temp has stabilized. During transition, either up or down, the in-socket thermistor is a few degrees behind the on-die thermistor, due to thermal propagation delay through the CPU ceramic. Lastly, make sure the Delta fan is blowing down into the HS, not sucking out. It makes a difference on the MC-462.
As for testing open vs closed case. When I am characterizing the performance of an HSF, I test with the case open, since I am not interested in characterizing the cases performance. When I am characterizing the entire systems cooling, I run case closed because that is how it will be used.
Hoot