Power in LTSpice

A quick note of something I found tonight.  I've been playing with some IRF510 designs, and for one of them, it looked like I was getting crazy good efficiency.  Like, not possible good.  I was evaluating this by running a transient simulation and then option-clicking (in osx) on the load resistor and on the IRF510 to get the power on those components, and then control-clicking on the labels in the graph to see the average power.  The power dissipation on the IRF510 was *way* too low (like 290mw when the output overall was 3.8w, which would be like 93%).  Turns out that when I was doing this, the current was negative some of the time, which was showing I guess as negative power.  So the average was much lower.  Wacky.  I guess I'm still learning this program.