The original poster shared some of the citations with me privately which helps a great deal in identifying what’s going on.
All of the original poster’s citations in RM7 that won’t merge after import into RM10 are using the Free Form source template and they all leave the Page Number field blank. There is nothing at all wrong with either procedure, and I have a lot of citations in my own database that are set up the same way. As the saying goes, “I resemble this problem”.
These sources are what are called split sources, or sometimes they are called extremely split sources. That’s because 100% of the source data is in the Master Source and 0% of the source data is in the Source Details. And when using the Free Form source template, there is basically only one Master source field, namely the Footnote fields, and there is basically only one Source Details field, namely the Page Number. Well, there are also Master Source fields for the Short Footnote and for the Bibliography, but those two additional fields have no impact on this discussion. What matters is that leaving the Footnote field completely blank is what creates extremely split sources when you are using the Free Form Source template.
Pretty much all the sources in my database are extremely split sources. Most of my sources are now based on source templates of my own design. But some of them are still based on the free form template because I have not yet gotten around to converting them from free form templates to using my own templates. With sources based on the free form template or based on the templates of my own design, if I were doing my conversion from RM7 to RM10 today instead of when I originally did it, I would not be able to do the conversion because of this problem with extremely split sources.
It’s probably worth reviewing that I became an extreme source splitter in RM7 only reluctantly and only because RM7 did not support reusable citations. Suppose I had created a citation in RM7 and then memorized it and pasted it many different places. Then suppose I found a typo or some other error in the citation. Because RM didn’t support reusable citations, I would have to chase down every single place I had pasted the citation that had the error and make the same correction every single place. What a pain that was.
It’s also worth reviewing that in addition to adding support for reusable citations, RM8 made a subtle but radical change in the way it managed the printing of citations in reports. In RM7, when you used the report option to combine duplicate endnotes, it combined them based on their text being the same. In RM8/9/10, when you use the report option to combine duplicate endnotes, it combines them based on the citations being the same citation that has been reused. As a practical matter, what this means is as follows. If you are an extreme source splitter who is converting from RM7 to RM8/9/10 and if you wish to combine duplicate endnotes in reports, then you are forced to run the Merge All Duplicate Citations tool. Otherwise, your endnotes will not combine. This tool worked for extremely split citations in RM8 and RM9, but it no longer works in RM10.
It’s worth reviewing why the change was made in RM10. It was made because RM8 and RM9 combined citations downloaded from Ancestry via TreeShare that actually were different but that did not differ in their text. Rather, they differed only in their media file or their Web tags. Therefore, such citations were merged even though they were really different. And that’s why the change was made to RM10.
In my personal opinion, the design of the change in RM10 is completely wrong. It still does not test for differences in media files or Web tags when merging duplicate citations, and instead it simply refuses to merge duplicate citations when the citation name is blank. That leaves all extreme source splitters in a lurch.
I therefore respectfully request that the recent change to RM10 be re-examined. The Merge All Duplicate Citations tool actually should merge citations with a blank citation name. And it actually should not merge citations that differ in a different media file name or different Web tag.
If I were converting to RM10 today, I suspect I actually would succeed. But I would have to do it the same way the Original Poster did. I would use SQLite to add some sort of dummy Citation Name to all my citations with blank citation names. Then I would run the Merge All Duplicate Citations tool, all would be well. But this solution is not available to most RM users, and it is not recommended by RM in any case.