Very interesting Jim. I don't have a spectrum analyzer here. I wonder if you could rig your Type II PHSNA to eliminate the ERA-3+ per my writeup in the K1RF folder entitled "Bandpass Filter Measurement Anomalies R2.pdf" I think you can greatly reduce the harmonics and still get about -8.7dBm output by:
1. Driving the antialias filter directly from the push-pull transformer with no attenuator per the AD9851 data sheet Page 11.
2. Take the output of the antialias filter to drive a 6dB attenuator.
3. Take the final output from this 6dB attenuator.
Then you are getting all your output from the AD9851 (or Type II AD9850) with no introduction of harmonics by the output amplifier. I just did not quantify the improvement but I see it in my measurements on some of the LC bandpass filters I designed.
If an output amplifier wants to be used, it really wants to be a low distortion type. However, any wideband amplifier also needs to be carefully laid out per the manufacturer's data sheet. But like one of my bosses has told me many times, "Keep it Simple Steve" so I prefer no amplifier at all if the output power is adequate. The The OPA2674 or OPA2677 with better thermal package would work if more output power is required at low distortion.
"Digital Steve", K1RF