Input data, target data and time difference _NARX

Hi, I am new to modeling by NN.I am using ntstool (NARX) for prediction. I have few question. If you kindly answer would be great.

1. Can I input matrix format of my input ( 5 inputs) and target ( 1 target) to the NARX ? I was confused by the input format example provided in NARX ( nsttool) in matlab that I attached here. Although my understanding is, I can input matrix form and tonndata will convert to cell array for NN. Please see the attachment.

2. I have attached Mydata_Matrix. In this file (yellow highlight-Head), I wont have any data from time 1 to 3 for Head ( Head data is not sensor data. This is coming with water to the station I am modeling. And would be reaching to my station at time step 4). So, I thought to delete my data rows from time 1 to 3 from my matrix (that will delete all elements associated with other columns in this 1 to 3) and start from time step 4. Is there any way to fill this value ? In my understanding, as I don't have data , I should delete. I have 70,000 data points( time steps) for my model.

3. For normalization ( zero mean, 1 std), do I need to delete outliers before normalization? As the outliers has significant impact on mean calculation? But, I want to keep all my data (including outliers), for my model training, validation and test. So, I was wondering, if I calculate the mean excluding outliers and apply (normalizing) that mean to all the data ( including outliers), will there be any significant issues?

Thanks