train custom NN with customised performance function using weighted error

Hi, I've created a custom neural network using a customised performance function which crashs Matlab everytime I try to simulate a result. anyway I've found that adding nn7 to the train's function input solves the problem:

net = train(net,x,t,nn7);

Now I can't find a way to add my weighted error vector ew to the train function, before adding the nn7 argument it was:

[net,tr] = train(net,x,t,{},{},ew);

I don't understand how the train inputs are handled, is there another way to pass the weights to the neural network without ommiting nn7, thanks.