It's not too hard to reverse engineer a curve with a surprizingly little amount of information. A basic quadradic curve can be drawn with the assumption Head at BEP and a 1.28 rise to shutoff, which will catch a whole lot of pumps, and that can be refined with a few other bits here and there. Start with assuming a typical centrifugal curve class. The curve number itself already gives away that it's a 13" impeller and a 6" diameter discharge. The more extra bits if info you can find out, the better the final curve. Estimating flow by assuming 10 fps at suction and 22 fps at dischage, is one example. Bouncing tangential velocity and head calculated from that, off the rpm and impeller diameter and estimated flowrates, eventually the bits start falling in place.
What would you be doing, if you knew that you could not fail?