The unannotated gene (not localized to the genome sequence) gene known as "Elongation factor 4A (Ef4A)" (FBgn0043346) in FlyBase doesn't exist in reality, even though it's associated with three papers (FBrf0128367, FBrf0147055, FBrf0152035) in FlyBase.
Here's what I think happened:
1a. This 'gene' first appeared in TABLE 1 of FBrf0128367 (published in 2000) as "elongation factor 4A (elF-4A)".  Here, it is associated with "EP(2)2317", which is stated to map to "34A5-6", with an insertion site "Clot 8989, -259-bp of GH06371".
1b. FlyBase curation of this paper resulted in:
a) the creation a new gene called "Ef4A / Elongation factor 4a" (FBgn0043346);
b) the creation of a new allele called "Ef4AEP2317" (FBal0119492);
c) the renaming of the associated EP insertion to "P{EP}Ef4AEP2317" (FBti0011032).
2. The 'gene' next appears in TABLE 1 of FBrf0147055 (published 2002) as "Elongation factor 4A", and is again associated with "EP(2)2317", which is again reported as mapping to "34A5-A6".  These data were presumably replicating information from FBrf0128367 directly or via FlyBase.
3. Finally, 'Ef4A' was associated with FBrf0152035 (published 2002) because EP line "2317" (mapping to "34A5-6") is mentioned in their TABLE 1.  HOWEVER, here the gene "downstream of the EP element" is given as "CG9332".
There are several apparent mistakes here:
i) cDNA clone GH06371 (mentioned in FBrf0128367) maps just upstream of CG9932.  This gene is located at position 34A4-5, consistent with the given map position of the EP element.  Thus EP2317 should have originally been associated with the annotated CG9932 gene in FlyBase, not with a newly created, unannotated 'Ef4A' gene.
ii) Unlike the 'Ef4A' gene, the associated P{EP}Ef4AEP2317 insertion IS localized to the genome sequence in FlyBase, and is inserted just upstream of CG9932 based on the flanking sequence with GenBank accession AQ073486 (consistent with point (i) above).
iii) There is no evidence that CG9932 encodes an elongation factor (as suggested in FBrf0128367).
iii) Further, "elongation factor 4A" is not a recognized eukaryotic translation elongation factor.
iv) FBrf0152035 says that EP2317 maps to "34A5-6" and is upstream of "CG9332", but the CG9332 annotation (which has since been split into the CG31673 and CG31674 neighbouring annotations) resides at position 38F1.
Interpretation:
A different clone, GH03937, was previously associated with the EP2317 insertion (according to my personal notes).  This clone has since been shown to be chimeric: it's 5' sequence maps to CG9932 (position = 34A4-5), while its 3' sequence maps to the "eIF-4a / Eukaryotic initiation factor 4a" gene (position = 26B2).
However, the NCBI record for the entire cDNA clone (GenBank AF145621.1) is misleading. It's entitled "Drosophila melanogaster clone GH03937 eukaryotic initiation factor-4a (eIF-4a) mRNA, complete cds".  However, the nucleotide sequence is a chimera of CG9932 and eIF-4a, while the translated amino acid sequence is that of eIF-4a ONLY and the link-outs point to eIF-4a ONLY.
My best guess is that the authors of the original paper (FBrf0128367) found the erstwhile link between the EP2317 insertion and the GH03937 cDNA clone, and found that this was associated with "eukaryotic initiation factor-4a (eIF-4a)" from the NCBI report.  But then they substituted a lowercase 'l' for the uppercase 'I', and 'elongation' for 'initiation' in the gene symbol and name, respectively, that they reported in their TABLE 1.
The association between EP2317 and CG9332 in FBrf0152035 is presumably a typo: the authors must have meant "CG9932".  The legend to the TABLE 1 says that "Genes of putative coding regions downstream of the EP element were identified using the GADFLY database of the BDGP", which means they would be aware that EP2317 was inserted upstream of CG9932, and were not confused by the prior associations between EP2317 and the mysterious 'Ef4A' gene.
Consequences:
1. Merge the 'Ef4A' Gene Report with the 'CG9932' Gene Report and make EP2317 an allele of 'CG9932' rather than 'Ef4A', thereby moving all associated information currently under 'Ef4A' to 'CG9932'.
2. Retain 'CG9932' as the FlyBase symbol for this gene.  Calling it 'Ef4A / Elongation factor 4A' based on FBrf0128367 would be inaccurate and misleading.
3. Rename the insertion from 'P{EP}Ef4AEP2317' to 'P{EP}CG9932EP2317'.
4. Remove any data associated with the merged gene that is based on the false assertion that it encodes a translation elongation factor.  E.g. GO terms.
Steven Marygold
FlyBase
Department of Genetics, University of Cambridge.