I checked the triple resonant box for the broadband EOM this afternoon, and found it was healthy.
So I installed it again together with a power combiner and succeeded in locking the MC.
Since the box has a non-50 Ohm input impedance at 29.5 MHz, so it maybe needed to adjust the phase of the LO for the demodulation of the PDH signal.
A good thing is that now we are able to impose the other sidebands (i.e. 11 MHz and 55MHz) via the power combiner.