@Ivan Rodriguez: It's best to calibrate the device and to power up the device at a voltage that produces the best outcomes. We ran a number of calibration attempts on different parts and concluded that we got the best calibration results at around 1.6V. Anecdotally, I have found, when testing an assortment of parts I have on my desk, that it works best to keep the voltage as low as possible without everything failing due to the low voltage, which is 1.43V (in the conversation just above, Matt Guthaus got better results at a little higher voltage, 1.45V, although that result might be part-dependent).
Jeff just updated the code base yesterday to include a new routine that sets the voltage without powering down the system, so you can bring up the device at the voltage where it was calibrated, let it program the GPIOs at that voltage, and then you can change the voltage to whatever you want for testing.