I do not have the requested information, but.. (and this is maybe off-topic)

I know using the FeatureReader where possible allows the spatial filtering to be done before reading the data and that is usually performing a lot better.

i couldn't seem to make a workspace runner work in conjunction with the feature reader when i tried it.

Easiest way of getting the polygon from the main workspace into the child workspace:

Create an attribute from the geometry (GeometryExtractor) (GML or WKT).

Child workspace:

Use Creator (Null geometry).

Use GeometryReplacer (GML or WKT) getting the definition from a published parameter.

Pass the GML or WKT definition to the child workspace through the attribute --> published parameter.

In addendum to Eriks suggestion i would shun WKT/WKB if you intend spatial filtering in oracle because you can't pass or create the srid wich is required in oracle.

I prefer to construct Oracle SDO strings. (seems daunting but once you get the hang its easy). You can either use them in an Oracle sql caller to spatially query some tables or recreate the geometry by selecting from dual.

Come to think of it..then you wont need the feature reader at all! You can do the query in Oracle Spatial/Locator.

there is an article comparing feature reader to spatialrelator and merging, somewhere here in the fme knowledge database....

There is a blog post, but I think that is comparing FeatureReader to FeatureMerger and Joiner for doing attribute-related joins. We're just starting to create a Common GIS Operations tutorial which will cover this spatial filtering/joining content (I hope)