I think the difficulty in developing/using mortality improvement factors for GI is that GI mortality is likely very different from company to company (more so than underwritten) because things like variations in anti-selection probably highly overshadow any general improvements in mortality. Anti-selection is dependent upon how you market/distribute the product, your level of control over how it’s sold, the volume of business, where leads come from, etc. Maybe you could reflect meaningful improvements in the (far) out years, but probably not so much in the early years.
That said (while I haven’t looked), maybe you could get some general idea from limited underwriting tables that the SOA publishes, if you can find tables developed from different eras and compare. Another source could be your reinsurers.