We didn't use ferrite beads for the power supply pins on the AD9912 evaluation board. So, I don't have a specifc ferite bead recommendation. However, some of our high performance clocking parts do find ferrites beads are helpijng to reduce coupling between channels and supplies. The ferrite part number used there was BLM15X601SN1D (600 ohms) from Murata. So, I would start with that one first, if it were me.
In general, the PLL phase noise performance is improved by running the PFD faster verses slower. Running faster uses less multiiplication to reach the final system clock frequency. So, picking the faster 100MHz TCXO could improve phase noise performance. Also, increasing the slew rate is typically a good thing too. So clippled sine wave sounds like a positive for slew rate.