Does object-oriented domain analysis work?

The Rational Unified Process (RUP), the Rapid Object-Oriented Process for Embedded Systems (ROPES) and other OO approaches propose Use Case driven analysis as a mechanism for capturing requirements and deriving object models. B. Douglass' Real Time UML describes how "later analysis decomposes the system into [...] objects". Beyond discussion of a number of common object identification strategies, such as noun-phrase underlining and key concept identification, only lip-service is paid to what has been described as the Fundamental Difficulty (FD) of Object Oriented Domain Analysis (OODA). The Fundamental Difficulty was defined by Svetinovic et al. (at RE'05 in Paris) as the difficulty of identifying system domain concepts as Objects (which some might think pretty central to the OODA enterprise). A study of undergraduate projects, documented by Svetinovic using these techniques, found that object models of the same system often differed drastically in terms of concepts identified, while software concepts were often specified at inconsistent abstraction levels. These observations raise the concern that OODA may be incorrectly applied by many practitioners. Microsoft's Steve McConnell argues that most practitioners neither have the benefit of a Software Engineering education, nor do they have ready access to evaluations of the myriad of available tools and techniques