Actually, the "process" can and does work for those who use it correctly, but you have to manage it properly or it can get out of control fast. There are a lot of pitfalls, some of which are under the program manager's control, some of which are external. One difficultly in estimating costs is the nonrecurring engineeering associated with RDT&E. It's, at best, a rough estimate because you cannot predict all the issues that may arise so if you get within
+ 25%, you're pretty good. Industry tries to win competitions pressing the technologcal edge (so-called bleeding edge sometimes) so if they can't deliver as promised, delays ensue and every day a program isn't delivering on time adds to costs not to mention additon costs to correct deficiencies and retest the system.
The Army has had some spectacular failures in trying to introduce technology and probably takes the prize in how not to do it:
Sergeant York DIVAD,
RAH-66 Comanche,
XM2001 Crusader, and more recently, the
Airborne Common Sensor (ACS). The Jury is still out on the
Joint Common Missile (JCM)
The result was billions spent and nothing to show for it, but lessons learned.