The problem in this reasoning is that precision does not increase the spin from h/2 to h in the (x,z) plane. The spin is a vector proportional to an area, in fact proportional to the magnetic moment squared. Mills says that, the x,z projections of the precision vector of h gives 1/2h and 1/4h what is correct(see GUT-CP figure 1.25). The resulting total spin (x,z) is then 0.559016994374947( (1/2)2 + (1/4)2)1/2. An other problem is that Mills talks of 60 degrees, what is correct if you use spherical harmonics modeling but this is the opening angle of the precision cone, that circulates around the central axes middle = 30 degrees! In fact h(x) 1/4 is not a "stable value" it's the magnitude of the spin precision component x of (x,z) in average it is "0" as spin does not add dissipative energy to a particle - added flux is just compensating the magnetic field and if the field vanishes the the two opposing fields annihilate!
To get the maximal possible h you have to add all 3 dimensions. But as mentioned earlier. Only orthogonal spin values can be summed up! The mistake Mills made is to either to assume that the magnitude of the S vector is constant h what is wrong or that S is circulating at a constant angle... or else an phi=45 degrees S would be > 1...
If things start to precess I would expect torque free precession. But I can't see that that is used. In recent versions of GUTCP there is a section that shows why the 60 degrees is needed. It is balansing of torques.