The ILO is very wide tolerance. A way around this would be a production test/cal routine,
where a known accurate source is fed to PSOC 4, it measures, and computes a G factor to
correct readings. But then to do that right one would have to sweep locus of T and V to have
complete correction, probably impractical. Maybe a first order correction might suffice, you would
have to work the numbers.
The TRM pretty much confirms what can be used as wakeup sources -