This is the answer I would give:

• The magnitude is lower because the differential noise of the input optics is suppressed by the secondary (CW) loop gain in the PLL readout (but not in the AOM readout). This was the main advantage to this readout scheme in the first place (see this post)
• The shape above ~100 Hz is different either because the residual primary loop noise ("spillover noise") is also suppressed in the PLL readout OR because the PLL bandwidth is lower than this. We don't have an OLTF of the PLL yet, but we can see which one it is by carefully reviewing the diagram in the above post to see if the spillover noise should indeed be suppressed (we didn't think it would).