I think the word you are thinking of is "monotonic".
It is very common to see bumps/spikes in a phase noise measurement. That bump at 3 Khz, is likely due to a pll control loop closing out at an open loop frequency of around 3 KHz. When a control loop loses its open loop gain, there is often closed loop peaking of noise at that point due to insufficient phase margin.
The transfer function of a simple closed loop system is G(s) = G(s)/(1 + GH(S))
when |GH| = around 1, and angle(GH) is around 180 degrees, you can see that G(s) blows up.
If you can not see the bump at 633 mhz, but it is there at 3x the frequency, remember that that bump will be 10 dB lower at the input frequency, so your measurment system may hit a noise floor.
It is also not that uncommon for a X3 multiplier to be have some inherent instabilities too. if it is an active multiplier, there could be a noise bump in the active device. If it is a passive diode multiplier, there could be parametric oscillations.