The DHT sensor series are not that accurate in humidity (~5%) and I do not know how the internal sensor works, but if some R.C. is at heart it might take more time to reach some voltage causing a different value.
How long is your cable. Just found this on the datasheet.
When using a 3.5V voltage supply cable length shall not be greater than 20cm. Otherwise, the line voltage drop will cause the sensor power supply shortage, caused by measurement error.
Thanks for the indication. The cable was ~30cm between the sensor and the board. I reduced it to <20 (~15cm) but the results remain similar, showing a msignificant gap between 5V and 3.3V reading
I'm using the AdaFruit's DHT library example code for this experiment. There is a 2 seconds wait between reads. Increasing it to 5 seconds doesn't change the outcome.