Confused about the equation just before Eq.18

[youwyu]

That's the equation just after "rewriting (6) and (7) as".
The △R_t1_t is not noise free, so I have two questions:

1. I'm not such an expert in Gaussian Process, so I wonder if △R_t1_t is a GP itself, and △R_t1_t(f-b_f) (where △R_t1_t is the ground-true rotation) is also a GP, how does △R_t1_t(f-b_f) obey Gaussian distribution?
2. If it obeys Gaussian distribution, then the variance in Eq.18 should not be σ_fi. It should take the error in Eq.17 into account.