The GDP mis-mapped the insertion site in MI02948.  When we submitted the data for it in 2011, we reported that it mapped to a site in the Release 5 3LHet scaffold.  The short 3' flank and the portion of the 5' flank abutting the MiMIC are repetitive and align with the consensus for TE roo.  In release 5, the alignment to this copy of TE roo in the 3LHet scaffold was better than at other sites.  However, in release 6, there are alignments of the repetitive segments of the flanks to multiple sites in the genome that are not very different.  We didn't catch this when we realigned all of the flanks to release 6 in the summer of 2014.
FlyBase currently has MI02948 mapping to  3L:24 ,858,927..24,858,927 -. However, the segment 1-153 of the 312 bp 5' flank of MI02948 is unique in the release 6 reference genome.  Our interpretation is that MI02948 is inserted in a copy of TE roo that is inserted in an intron of the tara gene on 3R.  The insertion site based on the unique segment is  3R:16233267..16233267  (+).  While it's on a different arm than we originally reported, it's on the same chromosome so the balancing was presumably OK.