The flanking sequence for the Exelixis c05541 insertion was never submitted to GenBank; the untrimmed sequence is included below (the lower case letters are lower quality than the upper case letters). When I BLAST the flank vs the reference genomic sequence, I get a convincing unique alignment of the flank 3-143 to  3L:1794766-1794915  (this range is within the dlt and alpha-Spec transcription units).  However, further distal in the sequence, there is a unique alignment of the flank segment 133-363 to  3L:3858054-3857827  (this range is within the Awh transcription unit).  This alignment ends with CCGG, which is probably the recognition sequence for the restriction enzyme used for the iPCR.
I can't really explain why different segments of the flank should map to very different sites on 3L.  Possibly there was an artifactual ligation during the iPCR.  It's also possible that there were two insertions or sample cross-contamination during the iPCR.
It's impossible to resolve the discrepancy without repeating the flank sequencing or in some other way remapping the insertion.  I would tend to believe that the dlt/alpha-Spec mapping is more likely to be correct, because this is the segment adjacent to the 5' end of the vector.
>PelementDBLane|PSeq99721|c05541-5 (untrimmed)
gaaggggggggggggagantttannnngggccttagttttcaggaagcag
cttcattttacgcagactatctttctaggttaatatcttgctaaccatac
agcaaagaagGGAGAGTGTGCTCTCCCGCTTTCTCTCATCTGAAGTTGCC
GCACCACCACCCGATGACGCACACACACACAGCGatctctactcctgcgt
aaccgcacacacacatgcacaagcatcagcggtaggcagtggtcgtgcca
ggagcatccgccgaccttcaggaagatccgatccgagatgggttccccgc
acgctgcgcatgaacgcatttaagtctggaaaagagtgggaatataaagc
gaaactcatggtcaccattgatgagcgggacaagctttctatgcaatata
ctggcgatttctttccattattcactcagacctgcaaaccggttttgata
tttctctgtgtgatttcttatgtacaagcgggcatattttgttttttttt
gctcccatttttccctttggctactttttgtttacaagtgttctcatatt
tttacatagttaaacataggtccacatcgcccccaccnggtccacttttt
ttttatttttgttctcttttttctactcgccgcccattttttattttttt
gttttattacgtcctacctttgcctcctttcgatttcctacgtaatgttt
tttctagttcttacaaacttcctttttatactcttctacgatcgagagac
gcctgtggacatacctttttgtttcttatagtttattactttttgcacct
aacagtctacactaactcccacntttaatattttttatgtttctttattt
acatatatatttggatactttgatgttaaac