So since your question is "what did you miss in your experiment", a few things I can think of:
- The blanket seems to be folded which is not the real application of the blanket. Stacking multiple layers may or may not change the results. I have no idea, but the blankets are not 100% reflective, i.E. some IR gets through the first layer.
- Use a proper setup. Strictly speaking, what you measured is the "current surface temperature (within measurement error of you IR cam) of a folded blanket". For all we know, you just held the gold side in your hand for a couple of minutes and it heated up more than the other side. Reflections from other objects nearby can also disturb the measurement. This is a known issue with IR cameras.
- There is more to reflexion than IR. Visible and UV light also contribute to heating.
- Usage might be dependant on emission/absorption and temperature of the human body compared to the surrounding environment.
Also I'm not sure if reflection is actually the proper measurement here. I'm thinking, you basically want to stick a thermometer in the blanket with the gold/silver side out and leave it in the sun/freezer for an hour or so. Which is more or less what the people from this paper did.
Now, to the second part of the question: Which side up? The results from the paper (and a few other sources across the web) basically say: It really doesn't matter. Reflectiveness between gold and silver varies by just a couple % and depends on the manufacturer.
it does not matter whether silver is turned down or not. There are
several factors to be considered, including convective and conductive
effects, vapor barrier function, and distance and moisture between the
body and the blanket
This is in line with my personal experiences when given an emergency blanket after cold races. The main benefit in those cases is that you can wrap the blanket airtight around yourself and reduce convection. It doesn't magically turn hot or cold when you use the "wrong side".