NARX how do feedback delays work?

I need to model y(t) using x(t-12) and y(t-12) only, hence have 'inputDelays = 12' but I cannot figure out whether feedbackDelays should be '12' or '1:12'. If the feedback is using the real target values y(t) I think it should be 'feedbackDelays = 12', but if the feedback is using the model output yhat(t) should I use 'feedbackDelays = 1:12' to produce a better fitting model? Would this adhere to modelling y(t) using x(t-12) and y(t-12) only?

My code is:

target= load('target_data.txt'); 

input = load('input_data.txt'); 

X = tonndata(input,false,false);
T = tonndata(target,false,false);

trainFcn = 'trainlm';  % Levenberg-Marquardt backpropagation.

% Create a Nonlinear Autoregressive Network with External Input
inputDelays = 12;
feedbackDelays = 1:12;
hiddenLayerSize = 10;

net = narxnet(inputDelays,feedbackDelays,hiddenLayerSize,'open',trainFcn);

[x,xi,ai,t] = preparets(net,X,{},T);

net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;

% Train the Network
[net,tr] = train(net,x,t,xi,ai);

% Test the Network
y = net(x,xi,ai);
e = gsubtract(t,y);
performance = perform(net,t,y)

netc = closeloop(net);
netc.name = [net.name ' - Closed Loop'];
view(netc)
[xc,xic,aic,tc] = preparets(netc,X,{},T);
yc = netc(xc,xic,aic);
closedLoopPerformance = perform(net,tc,yc)

nets = removedelay(net);
nets.name = [net.name ' - Predict One Step Ahead'];
view(nets)
[xs,xis,ais,ts] = preparets(nets,X,{},T);
ys = nets(xs,xis,ais);
stepAheadPerformance = perform(nets,ts,ys)

Thanks in advance, Alana