Like all tools, there's a bit of a learning curve. However the user guide does have quite a number of samples that show you how things are done.
There are two routes that you can take to process your data; use either a pure XML approach with XSLT to wrap your tags with the appropriate Miramo commands or use their macro pre-processor (mmpp) language (sort of like VBscript) to use your own tag structure to wrap the commands.
The workflow is typically such that one uses the database reporing tool to create the preliminary tagged output (much the same as you're doing now for creating MML text files). Then you use either XSLT or mmpp (though you have to write your own code for these) to process the output to wrap the actual Miramo commands around the content. Then you run Miramo on this wrapped content. Miramo uses your specified FM templates to get the formatting info and produces MIF as output that feeds back into FM to create the final book (and also creates all required generated FM files like the TOC and Index). Miramo also lets you call in scripts (Framescript or Extendscript) or other special APIs to process the FM content at specified steps in the book creation workflow.
Miramo is an enterprise-level tool that is designed for producing huge volumes of content. The Personal Edition is just limited in the number of concurrent output channels that be processed and requires a user to initiate each session rather than having sit on a server and automatically process content as it appears in specified locations.
The latest version also let's you directly process DITA content (using FM templates, you get more cosmetic and easier control over the output than you could using the DITA-OT) and also has a new module that produces PDF directly without even going through FM.
AdChoices