The simplest solution might be to create a single document that contains all the letter variants (or, at least the content that differs), with each variant given a unique bookmark name - preferably a name that matches the output from a field in the datasource that is used to discriminate between the letters, then use an INCLUDETEXT field in the mailmerge main document with a suitable MEREGEFILED in the bookmark reference position. For example:
Code:
{QUOTE{INCLUDETEXT "{FILENAME \p}/../Variations Document.docx" "{MERGEFIELD Variant}"}}
With the above construction, the document containing the variants would be named 'Variations Document' and stored in the same folder as the mailmerge main document. That document can even contain its own mergefields, if needs be. The {MERGEFIELD Variant} field would simply tell the INCLUDETEXT field which bookmark to look at, according to whatever;s stored in a data field named 'Variant' in the data source.