A comment by Murugappan Ramanathan, over in the Bizagi professionals group (this is an adaption of my comment there) started me thinking (rather in this context) that one should always model what a process should be like and look at the gaps between that and what the process is and what cost-effective automation can support. That said, a lot of process doesn't really differentiate a company and you might as well adopt whatever is easiest, using SAP, say, as an example of general "good practice". That will save money and resources, which one can devote to automating and maintaining processes that really differentiate the company and make it competitive.
A Bizagi BPMS business model - or any other business model that incorporates completeness and consistency checking, etc - might be a great aid to identifying the unique processes that give a firm its competitive edge and differentiating them from processes that might as well go into SAP and run in whatever way best suits SAP. This is equivalent to "keeping the BPM separate from the main apps" in my view.
My caveat is that these must be fact-based choices, not based on "not invented here". SAP I can quite believe is sometimes rigid and "opinionated" - at least from some points of view - but it does also represent more than adequate "good practice". Many firms think that what they do is different to whatever anyone else does - but, below the cosmetic level, they are mostly wrong (everybody's sales invoicing process is probably much the same, disregarding cosmetics, and one might as well use whatever SAP provides). The skill is to understand both the logical business model and the physical automation model (possibly implemented in several technologies), and prioritise addressing the gaps between them in terms of their contribution to competitiveness.