This paper addresses the initial shift problem in iterative learning control with system relative degree. The tracking error caused by nonzero initial shift is detected when applying a conventional learning algorithm. Finite initial rectifying action is introduced in the learning algorithm and is shown effective in the improvement of tracking performance, in particular robustness with respect to variable initial shifts. The uniform convergence of the output trajectory to a desired one jointed smoothly with a specified transient trajectory from the starting position is ensured in the presence of fixed initial shift.