They could all be within specification, except for one (t4). From the data sheet:
“±0.5°C Accuracy from -10°C to +85°C”
so the spread from the lowest (21.8°C) to the highest below t4 (t11 at 22.7°C) is within that 1 degree spread. 85 °C (t6) is a fault - it was asked for the temperature but it hadn’t seen the command to “convert” first. This could be a loose wire.
Take a look here, if you haven’t already: Ds18b20 and emonTX3CM firmware. It’s a long thread. Down at post no.65, there is 'scope picture of the timings for reading two sensors, but using emonLibCM, NOT using the emonPi as you have it - I think it will take that amount of time - 375 ms with 11-bit resolution) for each of your 14 sensors.
And further down at post no.78, there are timings for a whole range of “cloned” sensors. If one sensor is always significantly different to the others, check its serial number against that list.
If you found you needed to repeat a command two or three times to get it to take effect, that tells me there’s something not right with the procedure you are following. Advising others to follow that procedure isn’t likely to leave you in the best light.