merge but by group

You can use a FeatureMerger to join both datasets on the category attribute, make sure to create a list and then use a ListExploder to turn that into individual features again since there's a one-to-many relationship.

Then you can set a fanout on the Excel writer, using the client attribute as the filename, so for each unique client name you will get a new Excel file.

I've attached a template with this workspace and the sample source data:

none2none.fmwt

If the above should in fact include EE621984 then I would aggregate your first table, Group By Category, and create a list called CLIENT and join this to the supplier on a featuremerger

The second table would then be joined to the requestor port and merge on the CATEGORY attribute, after merging use a LIST exploder which will create a feature per cell per relevant client

Depending on the source format (and from the word table I take it is a database), you could also use the SQLCreator and have the database do the joining:

Select * from Client_table t1, Cell_table t2 Where t1.client_code = t2.client_code.

Then write the Cell_name and use the Client_name as attribute to name the output file.

Another thought. The InlineQuerier (SQL statement) could be used here regardless of the source format.

join-tables-with-inlinequerier-example.fmwt (FME 2017.0)