The only thing I would debate with him is whether we should go "broad or deep." That is, do deep discovery of all the intricacies of a selected few high impact applications - or do comprehensive, but more shallow discovery across the entire infrastructure. Deep is good for high value service management - but if you want to do chargeback, portfolio management, massive server virtualization, or other broad stroke efforts, you need complete visibility across the data center at least for the basic app/host dependencies.
Notice that he also has little use for the vague term "IT Service," focusing on applications and the application boundary problem.
He also touches on the perhaps problematic relationship between BSM and CMDB. They overlap, but are not necessarily the same thing. The CMDB implies a comprehensive scope, but true BSM (instrumented service availability management, using dependency maps) may only have ROI on a subset of applications in the data center.