Ferrite might at least give you a sensitivity / response
clue.
A question is, whether the emissions are current mode
(magnetic) or voltage mode sourced. For example a
fast diode with low series resistance will keep the
voltage overshoot down, but maximize current, and
the low series R increases the Q of "tanks" that involve
the diode. You may be seeing "tone bursts" that come
from ringing, as your EMI bad boy. Looking at the
board with a "sniffer" on the spectrum analyzer might
help localize the source, and using a "sniffer" that
does not change node C, with a 'scope, looking for
the same frequency (maybe not continuous, but look
for damped ringing cycles that just match) could put
you close to the source (process of elimination).
Here's a full bandwidth 'scope grab from one of my
CMOS POL DC-DC designs, a hard switching buck
with an EVK that used the lowest-ESR caps we could
find (and plenty of 'em). Note the harsh HF on the
magenta trace which is VIN. There's probably an ideal
Q for the input filter and importantly the VIN-GND
decoupling (a prime "tank" recirculating GND perturbations
back to VIN).
With caps, and I suppose diodes, "more ideal" may
not be better for EMI. Efficiency, maybe.