If you have a multimeter that can do 4 wire temp measurements, a PT100 is probably the most precise a consumer can get at room temp. It automatically compensates for wire resistance and such. Standard for measuring temperatures in a smaller range than a thermocouple and doesn't need as much amplification.
The problem is that there is not a certainty that the SHT and DS18 are just wrong by offset. Looking at your graph there is quite some unstable measurement accuracy in one or both of them too. The top line has many sustained spikes that the bottom doesn't, for example.