It appears that the inverse search will not work properly
in documents with two (and presumably more) columns.
Since the referencing information about the source file
and line number always applies for the whole text width
(it spans both columns), the destinations from the
columns mix together. This is especially annoying if the
source files are different for the two columns.