I have some trouble to understand the proof in Nielsen&Chuang about Knill-Laflamme conditions.
The conditions:
Let $C$ be a quantum code and $P$ the projector onto $C$. Suppose $\mathcal{E}$ is a quantum operatiion with operation elements $\{E_i\}$. A necessary and sufficient condition for the existence of an error-correction operation $\mathcal{R}$ correcting $\mathcal{E}$ on $C$ is that:
$$P E_i^{\dagger} E_j P = \alpha_{ij} P$$
For some Hermitian matrix $\alpha$ of complex numbers.
To show the sufficient condition, he diagonalizes the matrix $\alpha$. Doing this, we end up with a simpler condition:
$$ P F_k^{\dagger} F_l P = d_{kl} P $$
Where $d_{kl}$ is the matrix element of a diagonal matrix.
The polar decomposition tells us that there exist a unitary $U_k$ such that:
$$F_k P = U_k \sqrt{P F_k^{\dagger} F_k P} = \sqrt{d_{kk}} U_k \sqrt{P} = \sqrt{d_{kk}} U_k P \tag{1}$$
My question
At this point, he says that the effect of $F_k$ is therefore to rotate the coding subspace into the subspace defined by the projector:
$$P_k \equiv U_k P U_k^{\dagger}$$
I am not sure to understand precisely this statement. I agree that $P_k F_k P = F_k P$ such that $P_k$ stabilizes the space $F_k C$. But how to be sure that $P_k$ is exactly a good projector. What I mean is that veryfing $P_k F_k P = F_k P$ is for sure not enough as $P_k=I$ would also do the job for instance.