One way would be with conditionals and adjacent feature handling.

This workflow should do what you want

The AttributeRemover and AttributeKeeper aren't strictly necessary as you could set Conflict Resolution to Use Supplier in the FeatureMerger instead

@pedro I've attached the example workspace to illustrate @jdh 's solution (2019.1): adjacentattributes.fmwt

More on the Adjacent Feature Handling in the AttributeCreator documentation under the section Advanced Example: Using Multiple Feature Attribute Support

I would similarly use essentially the same solution shown by @ebygomm. The key is DuplicateFilter, what it does is take the "first" feature within a Group and sends that to the Unique Port, and any others within that group gets sent to the Duplicate Port.

If you send a Sorted table, then DuplicateFilter will automatically then take the Top/First result in each Name group, and then the Top/First Color can be joined back into the original table with a FeatureJoiner or FeatureMerger (FeatureJoiner tends to work faster for large datasets).

Alternatively InlineQuerier can do it in one step. If you send the unsorted, raw table to InlineQuerier and say name the table "ColorTable" then using this as the SQL Expression for the Output within the Transformer will work:

SELECT A.Name,
(SELECT Color FROM ColorTable WHERE Name=A.Name ORDER BY Area Desc LIMIT 1) AS MaxColor,
A.Area
FROM ColorTable A

Essentially it makes an Attribute "MaxColor" as a value derived from Subquery that sorts by Area in Descending order within its Name Group, and takes the top row as the Color result (this is what "LIMIT" does)

Input Table

Result Table