Hello,
I think the basic requirement for an accurate capacitive level sensor is a homogene and well defined medium to be sensed. As far as I know, capacitive level sensors are commonly used for liquids and sometimes for bulk material. Accuracy could be expected mainly for liquids.
A ideal medium is probably water due to its high er. You can try to manage, that you measure mainly the well defined capacitance of your sensor electrode isolation, e. g. a PTFE tube. With nonconductive liquids, dielectric properties contribute considerably to measurent results and should be expected temperature dependant.
I don't know if there is a principal behaviour in this regard, but probably polar and non-polar substances behave different. For pure compunds, you can hope to find some dielectric data in the internet, e. g. from the NIST chemistry webbook.
https://webbook.nist.gov/chemistry/
You could try to find an electrode configuration, that allows to measure medium properties separately, or at best, eleminates its influence by measuring a ratio of two signals.
Precision industrial level measurement has turned to use radar like techniques with open transmission lines because the limitations of capacitive measurement.
Regards,
Frank