The principle is that the disabled person should not be exposed to a significantly greater fire risk than any of the other building occupiers. Therefore, if you can reasonably argue that they are either able to self-evacuate at the same time as the rest of the occupiers, or that there will be an assisted evacuation out of the risk area to a refuge or place of safety either with or at the tail-end of the evacuation of everyone else, then I'd suggest that would be sufficient.
If the initial evacution is to a refuge, you then have to be able to credibly describe how they will reach a place of safety (with management assistance, if necessary), without relying upon assistance from the fire brigade etc. You have to be able to describe how this happens reasonably swiftly, & if there is likely to be any appreciable delay the disabled person should either be attended or have a means of communicating with those managing their evacuation. Such delays must be minimised, & assuming that the fact that you have (say) 30 minutes fire separation means you have 30 mins to evacuate them is not, of course, correct - evacuation from the refuge should be completed much more swiftly than this.
PHE is fine, so long as the fire protection is up to scratch & you can iterate how it'll be managed - effectively it uses the adjacent compartment as a large refuge, & the same comments as regards management & communication apply.
