Hi,
The exact location of the temperature sensor in the IC is not relevant; the bottom side has the largest thermal conductance. Temperature changes will be picked up quicker via the bottom - but the top side surely also contributes significantly! The graph you posted is not unexpected: the material used for the encasing and the amount of air surrounding the IC influence this behavior a lot.
I am assuming you are using the HVQFN24 IC, mounted on a PCB. A typical PCB is not an efficient conductor of heat. Regarding the package, both the WLCSP25 and the W8 package will behave better in this regard. You can also try to use another type of glue or physical connector to better guide the heat towards the IC.
From the graph I don't get to see a timeline - I don't know your measurement interval. There is no point in optimizing beyond the needs of your use case - perhaps the slower temperature acceptance is acceptable after a few tweaks in your material?
You can search around for thermal conductivity in general; there are also specialized firms that can provide you with professional support in this - I'm not a physicist nor a material expert (far from it!).
Good luck,
Dries.