R-Squared Difference when Dependent variable is $/lb vs Total Dollars?

[stats1322]

Why would I see a fairly significant difference in r-square (98.5% vs 85%) when I change my dependent variable from Total $'s to $'s per lb? My independent variable in both cases is weight. Each equation yields essentially the same answer in all tests of weight (within .08%).

How do I know which one best represents my data? It makes a difference when talking to management if something explains all but 1.5% of the variation in price, vs 15%.

 

[stats1322]

Example of the difference...

A part that weighs 2 pounds:
Total dollars equation estimates $72.645
$/lb equation estimates $36.33/lb (36.33*2 = $72.66)

This is how it looks for all the different parts.