With 40 years of data you should be able to estimate the 100 year flow with a reasonable degree of certainty. If you don’t have enough observed data for calibration, you do the best you can and clearly and emphatically state the analysis limitations; maybe present reasonable bounds of your results. You don’t need a rainfall-runoff model, but it certainly seems like it might help you if you have sufficient data. The new version of HEC-HMS has some very powerful uncertainty analysis features you may find helpful (User’s Manual Chapter 16).
An important thing to think about is what the “100 year event” means for your analysis. Is this the regulated discharge from the system, or the “natural” discharge of the system were the dams not in place? One of those will require a lot more work than the other.