From this Question: ESRI Shape files with UTF-8 encoded field names ?
The issue is more serious in FME Cloud (FME Server/Engine on Linux).
Most users in Japan use Shapefile datasets with the cp932 (Shift JIS - system encoding in Japanese Windows). However, since FME Cloud always treats attribute names of Shapefile datasets with UTF-8 (system encoding in Linux), all the datasets with Japanese attribute names cannot be processed with FME Cloud. MapInfo (MITAB) seems to have similar issue too.
I'm afraid that the problem above might be the tip of the iceberg - i18n issues on Linux FME engine.