confused by "predict" fuction!

Here is the link for explaining the difference between k-step ahead prediction and simulation performance of a linear model:

http://www.mathworks.nl/help/ident/ug/simulating-and-predicting-model-output.html

I have the following code:

-----------------------------------

u=[10; -20; 50; 70; 30; 90; 0; 120; -60; 40; 80; 40; -60; 10; 130];

t=[1:length(u)]';

y_meas_0 = 80; % y0 for real system

y_mod_0 = 100; % y0 for approximate model of real system

y_meas(1)=y_meas_0;

y_mod(1)=y_mod_0;

for k=1:length(u)

    y_meas(k+1,1)=0.95*y_meas(k)+1.45*u(k)+sin(k); % real system response

    y_mod(k+1,1)= 0.9*y_mod(k)+1.5*u(k);        %  model response

end

%M = idss(A,B,C,D,K,X0,Ts)

sys_mod=idss(0.9, 1.5, 1, 0, 0, y_mod_0, 1);

z_meas=iddata(y_meas(1:length(u)), u, 1); % measured input-output data

opt = predictOptions('InitialCondition', y_mod_0);

[y_pr, x0e, sys_pred]=predict(sys_mod, z_meas, 1, opt); % 1-step prediction

y_sim=lsim(sys_mod,u,t, y_mod_0); % simulating model

comp=[y_meas(1:length(u)) y_pr.y y_sim(1:length(u))] % comparing results

-----------------------------------

However, I see that the predicted outputs are the same results as simulated outputs. This is opposite to the explanation made in the above link. I am confused. I will be happy for expert comments.