Why the optimal value of variable obtained by cvx isn't satisfied the constraints?

cvx_begin sdp
variable Theta_W(N,N) hermitian semidefinite;
expression F_T(K,1);
subject to
for i = 1:K
F_T(i) = real(trace(F(:,:,i)*Theta_W));
end
for j = 1:K
P * (F_T(j) - Q * sum(F_T(j+1:K,1))) >= Q * sigma2;
end
for k = 1:(K-1)
F_T(k) >= F_T(k+1);
end
for n = 1:N
Theta_W(n,n) == 1;
end
cvx_end
This is my code, the value of optimal variable is not satisfied the constraints. Is there something wrong in code?