If you are referring to the close spaced lines
making up the spectrum of the two strong signals, from teh plot
you show they appear to be about 1.5Hz spacing. This is the
WSPR symbol rate (actually 1.465Hz) so the lines visible are
the natural sidebands from the WSPR waveform.
I'd be surprised if an DDS itself were generating 100Hz
sidebands on its own. Especially a more modern one like the
AD9859. Even the venerable AD9850 doesn't generate stuff
only 40dB down. So I'd still look for other means of 100Hz
leakage. Perhaps getting onto the reference source. I
assume the two stations both reporting 100Hz sidebands are
using identical generating hardware ?
I use an AD9852 for WSPR and generationg MFSK modes
generally, a device perhaps bit a bit older than the AD9859,
but the same family, and it is certainly clean down to at
least -60dBc. When driven from a higher clock frequency,
better still.
Andy G4JNT