Can't get NARNET to produce meaningful result on own dataset

Hi,

following the example for the narnet from the documentation, and predicting 30 more data points I get the following result which shows the last 30 data points from the original dataset the simulated values as well as the predicted values.

When switching to my own dataset (2933 samples) which similarly consists of values between [0.0277 0.9990] and running the same statements as before, the result I get does not fit at all, as shown in the plot below. In particular the predicted values are totally strange.

This is the code:

% X = simplenar_dataset; which = 'simplenar';
X = data; which = 'my';

net = narnet(1:2,10);

net.trainFcn = 'trainlm';

[Xs,Xi,Ai,Ts] = preparets(net,{},{},X);

[net, tr] = train(net,Xs,Ts,Xi,Ai);

[Ysim,Xf,Af] = net(Xs,Xi,Ai);

perf = perform(net,Ts,Ysim)

[netc,Xic,Aic] = closeloop(net,Xf,Af);

Ypred = netc(cell(0,30),Xic,Aic);

figure;
plot(cell2mat(X(end-30+1:end)),'r-o');
hold on
plot(cell2mat(Ysim(end-30+1:end)),'b-');
plot(cell2mat(Ypred),'g-');
title( [which ' dataset'] );
legend( 'real values', 'simulated values', 'predicted values' );

You'll find the data.mat attached to this message.

Any hint on what is going on here is greatly appreciated.

TIA Thomas.