Too fast, minaya!
I think I have found the problem here.
I was using digital 2 and 3 in SoftwareSerial (the old known NewSoftwareSerial) in my sketch. Maybe this pin3 (PWM) communication made the interferences I saw before in pins 5 and/or 6 (both PWM)?. Some preliminary test show that if I turn off the SoftwareSerial comunication, measures tend to be steady.
I'm trying now to use the turnOffPWM() I mencioned before to disable the PWM in pin3, maybe this will make the difference?