I have updated what I call my reference implementation of an XBRL-based public company financial filing to the SEC. You can get to the new reference implementation here. There you will find a high quality example of an XBRL-based digital financial report with complete human-readable and machine-readable documentation.
If you gave the exact same financial information about one economic entity to ten different professional accountants; you would expect that all ten accountants would create a fairly similar financial statement for that economic entity. Each professional accountant would represent that financial information similarly in a set of financial statements in all material respects. Each accountant would create a true and fair representation of the economic entity's financial information in the financial statement they created. That is true whether the information is formatted in HTML or PDF or Word or XBRL. Right?
Try that experiment with an XBRL-based financial report. How do you think that might turn out?
How would you define "correct" or "best modeling approach"? This is how I defined correct for my reference implementation:
If how to create a digital financial report is not well understood, then XBRL-based digital financial reports will never be usable. A digital alternative for a general purpose financial report will not magically appear. What a digital financial report ends up being will come from deliberate effort, skillful execution, and rigorous testing.
The documentation of the reference implementation will help software vendors understand digital financial reports. All the files are there for software vendors to use for testing.
Check out the reference implementation. Understand digital financial reports by analyzing and experimenting. It is through this analysis and experimentation that professional accountants will understand how to make digital financial reports work reliably and predictably.