A question about learning the EKF algorithm

I am learning the EKF algorithm.
I have seen the https://github.com/priseborough/InertialNav/blob/master/derivations/SymbolicOutput24.txt

as this article said, the predict state is X(k+1|k)= F* X(k|k),
so q0 = q0 - 0.5 * (dax * q1 + day * q2 + daz * q3) + q1 * dax_b + q2 * day_b+ q3 * daz_b;

but we all know q0 = q0 - 0.5 * (dax * q1 + day * q2 + daz * q3).

Does anyone know what is wrong about my understanding?

as the article said,
X(k+1|k)= F* X(k|k)

q0= q0+ SF(10)q1+ SF(12)q2+ SF(11)q3+ SF(15)dax_b+ SF(16)day_b+ SF(17)daz_b
= q0- 0.5
q1+ day
q2+ daz
q3)+ q1
dax_b+ q2
day_b+ q3*daz_b

I think I have get the reason.

‘F’ is used to calcaulte the covariance, and can not used to calcalute the predict state.
‘f’ is used to calculate predict state, and ‘F’ is getted from 'f’s jacobian.