Even in a model without air resistance and without any friction at the
axles, a fraction of the PE decrease has to go to rotational KE of the
wheels. As you increase the mass of the cart body (without changing the
wheels), the final translational KE constitutes a larger percentage of
the total KE. Put another way, a larger percentage of the PE goes to
translational KE when the cart body is more massive, resulting in a
faster cart. That may be why it pays to screw up your cart so that you
only have three wheels that need to be rotated instead of four.
At our pinewood derby, we had an extra "cheater's race", where anything
was allowed provided it didn't pose a threat to the facilities. Out of
curiosity I entered a Pasco dynamics cart (slightly modified to increase
the amount of clearance under the body). Its time was about a factor of
two faster than the average.