You have the math turned around. Because the nuclear resonance is much more stable and high frequency the Q factor and accuracy of the measurement is higher. With a cesium or rubidium clock it's very difficult to control all the influences on how tightly the nominal resonance is achieved and the Q while impressive is a bit less.
There are some real challenges in realization: this will take optical combs and all sorts of other stuff to really take advantage of.
They also point out that because the thorium atoms can be embedded in a solid, and have motion << the wavelength of the radiation, the emission and absorption are largely recoil-free. This eliminates Doppler broadening. What broadening there could be was below the resolution of their pump beam.
There are some real challenges in realization: this will take optical combs and all sorts of other stuff to really take advantage of.