There are two decoupling caps right at the chip on the underside of the pcb, I can't get them any closer without cutting the top off and mounting them on the die :) <-smiley face
Another one next to the V+ pin on the underside and the 10µF next to the other V+ pin.
Looking at the evaluation board, which has much longer supply lines, it only has one more than this around the DDS chip, the others are associated with the interface chips.
Re the supply to the xtal osc, the LP2985 is listed as a low noise regulator and I believe should be more than adequate.
It is well decoupled and again is on the underside of the pcb.
I did some digging through Analog's forums and engineering notes today and one thing that I noticed was that, in their own words, some of their recommendations for decoupling, seperate ground planes etc was somewhat over the top and simpler schemes made no noticeable difference.
Re oscillator phase noise, well you're already hindered by the limitations of the AD9850 to begin with, so I can't see using ultra pure regulators and massive amounts of decoupling are going to do anything useful. I could be wrong, which wouldn't be anything unusual, but I don't think so in this case. The obvious proviso is of course to use a good quality Xtal Osc module.
The test will be to build a couple and compare them against the standard module and rework it if necessary.