Unless these tests are done in an equal still air controlled ambient
environment throughout the test period, I would not trust the results.
A good test may show only three liquid ounces of ice melt difference
between all three boxes in a 14 hour test. If you consider a liquid
ounce consumed only about 10 Btu during the phase change, how important
would the three ounces be? Now the equation must be reduced to heat
loss by quart of air space in order to compensate for different box
sizes. If the delta T were increased by having all boxes in a
controlled environment of 110 degrees the ice melt might mean
something.

How were you purposing to calculate the changing box exterior
temperature as the day night temperature changed? And what margin of
error would you use plus or minis three ounces of water.