FYI, I just tested exporting the same DAE to GR2 using the option "Apply Basis Transformation to Y-Up" both checked and unchecked, and I see no apparent difference between them, the meshes are properly placed (unless this option only affects normals???)
More specifically, if the DAE you're importing has "<up_axis />" then it won't do anything.
If it has "<up_axis>Z_UP</up_axis>" or "<up_axis>X_UP</up_axis>", then it is converted to Y-up.
It is not specific to normals, it affects all properties of the model.
Then I misinterpreted what "Recalculate inverse world transforms" do.
IWT recalc is usually not necessary as the converter uses the IWT that is baked into the DAE when exporting. Sometimes this information may not be correct, in which case a recalculation is needed. But it is more of a debug option than something you'd actually need to use.
Inverse world transform is also known as inverse bind shape matrix / bind pose.
If I get a GR2 that is getting me the same issue again, I'll make sure to post it here, maybe then you can shed some light about it.
Please post the original DAE and corrupted GR2 file too, as apparently there is a specific configuration that causes this issue.
