Very good suggestion and one that we will get to. Probably not FME 2018 sadly but this is on the roadmap for certain. TO do this we have to bring feature type parameters into the feature reader, like they are for the feature writer. A necessary but big step in our plan to make FeatureReader the item of choice in the future.
This is kind of very basic reason not to read whole data sets for large number of fields , in some workflow we just need one or two fields from large data sets