Trying a different approach, from the circuit here (pdf file). This uses a 2n3053 to amplify an oscillator and drive an IRF510.
I built it out on a breadboard and after a lot of tinkering am getting about 0.7W out. Which is nice and all, but for the voltage I'm using I should be getting twice that at the voltage I'm using. Also, if I change out some of the coupling capacitors the plans use for some larger ones, it works even better in the simulator. However, when I try to do it on the breadboard, it causes the DDS to stop working. Not sure if there is some weird oscillations occurring there.
I think the big issues is that when I turn it on, the waveform going into the 2n3053 goes from a nice sine wave to something that has a lot of higher-frequency content. Not sure what the cause of that is. It could be just that the parasitics of the breadboard are causing it and that I should just go ahead and move it to copper, but I kind of wanted to get it working better before I did that. Ho hum.