Dear Jianhua,
unfortunately, to answer your question correctly, essential information is missing, you do not mention the reference sequence(s) used. We assume you mean variant NC_000014.9:g.45185212_45185215del NM_020937.2:c.4516-5_4516-2del.
For this description, indeed, the 3'rule needs to be applied and the deletion of one of two CTTA-copies from positions c.4516-9 to c.4516-2del is described as c.4516-5_4516-2del. and NOT as c.4516-9_4516-6del. Note that from deletion and duplication variant descriptions, since the 3'rule might be involved, it is not possible to deduce the exact site which is impacted by the variant. This includes especially deletions/duplication near splice sites. We therefore see no reason to change the HGVS nomenclature by adding the suggested exception.
Best regards,
Johan den Dunnen
HUGO HGVS Variant Nomenclature Committee (HVNC)