After three weeks of humidity measurement, Azure data shows the following behavior:

I expected to see straight lines for sensors blue and green as they are in test tubes.
The measurement is not random though because:
1 - The fully immerged sensor (blue) gives higher values than the half immerged (green).
2 - Green and blue follow exactly the same pattern
3 - Red sensor which is in a pot, shows similarities in its evolution.

I copied local temperatures history from the same period from the Web and compared the two graphs on a day per day basis:

It is obvious that my "sensor in a test tube" is a pretty good thermometer!! 
In order to know the humidity of my pots, I need to correlate sensors information with local temperature and add some filtering mechanisms.
I added a DHT11 Temperature and Humidity sensor to the Azure Watering system (another option would be to get it from the cloud...)