I reckon that's completely expected as this type of polymer humidity sensors are very rough and ready in their operation (especially without individual calibration). They are also history-sensitive especially if subjected to 100% RH.
For instance the SHT20 datasheet gives accuracy as +/-3% at a fixed temperature ignoring hysteresis and long-term drift
In other words you don't actually know the repeatability and accuracy as you cannot ignore hysteresis or drift in an actual application - or put simply the technology is pretty flakey if you want actual accuracy rather then a rough guide and trend-spotting.
You need a wet/dry bulb system to measure relative humidity with reasonable accuracy, or something really high tech using IR absorption spectroscopy.