As we can see from the stats im OMIM http://omim.org/statistics/entry the ratio gene:elucidated molecular defect with phenotype is 13553:3514.
The problem that I have seen in many databases, even in UniProt itself (but they manage 13,686 x-refs), is that links get put in (automatically or via curation), merely because they happen to have an OMIM entry but there may not be any molecular elucidated mendelian disorder actually associated with that gene. The reports are good summaries but the specificity of the links becomes useless in the sense that they are not actually selective for mendelian disorders, defeating the object of including them.
Is there any way to query these out via OMIM or some other source as a UniProt or EGID list for just the 3514 disease-causative molecular variants ?
just a comment on the name of the database: MIM was the name of a catalogue of mendelian disorders started in the 1970s, while OMIM is the current online version of the database.
so, basically you want to make a search in OMIM, but filtering out all the entires that do not have any elucidated molecular defect?
Yup, and your right on the name point. I also know it historically as OMIM even though UniProt renamed it to MIM in the x-ref lines