I found FME to not be very consistent reading MGA94 sources, in that similarly when first started to try to depend on the Reader / downstream Transformers to automatically determine the feature coordinate system from the Source Readers, would end up with 2 variations of the same FME coordinate system code for an identical actual coordinate system.

In contrast do not have the same problem with MGA2020.

The workaround in MGA94 data was to manually set each of the Readers/SQLExecutors etc. CoorindateSystem to enforce tagging of MGA94-xx to each feature read into the workspace rather than leave the Reader parameter blank to let the Reader come up with the Coordsys FME Code.

By setting this manually as a Reader override, you will get warnings that something like “Feature Type … Source says its Coordinate System is _MGA94-55_0 but you set the Reader to override this to MGA94-55 so am going to use MGA94-55 instead”.  But this way forces all workpace features when read in to be a consistent coordinate system rather than the jumble of MGA coordinate system codes can otherwise end up with.

I suspect the FeatureJoiner error getting though is for the optional setting of trying to create an Aggregate geometry out of Left and Right Input port features, so be careful that is indeed what you want.  I rarely use FeatureJoiner’s Geometry aggregation setting as just don’t have many use cases for it.