The Flow appears to be quite consistent though, depending on the mode in which it is used, it can be quite inaccurate.
Below is a plot of reported power for speed on the Flow, with each line representing a different "scale factor." All of these data were collected at a coast down calibration of 0, with the same tire, at the same ambient room temperature; however, the red dots represent data that were collected 4 months later than the black dots. As can be seen, when conditions are matched, the consistency of the Flow appears to be quite good.
The plot above compares the Flow to itself. Similar to what zenbiker has noted above for the Tacx Fortius, when compared against an on-bike power meter the Flow tends to underreport -- though as one might expect from the plot above it underreports consistently. In my own case, with my own tires, a better agreement between the Flow and a Power Tap (which had been statically-checked for accuracy) could be achieved with a coast down calibration of 4 and a scale factor of 120, though I suspect that could differ from tire to tire.
A greater problem for the Flow is its ability to adjust load while in "ergo" mode. In that case, one sets the power level and then, in theory at least, one can pedal at different cadences and speeds while the Flow adjusts the load to keep power constant. Sadly, here the Flow was overmatched. The plot below shows the percent error for three different scale factors (80, 100, and 120) when plotted against the ratio of "ergo power" to speed. That is, suppose you set the Flow to hold its power in ergo mode at 250 watts and pedaled at 20 km/h, then 30 km/h, then 40 km/h, and so on. The ratio of ergo power to speed (in km/h) would be 12.5, 8.3, 6.25, and so on. If the Flow were able to adjust its load at different wheel speeds, each group of points would be horizontal and parallel, and, in an ideal world one could adjust the scale factor to move some group of points up or down until it hovered over the zero error line. Why look at the ratio of power to speed? Power is the product of wheel speed and wheel torque (and a constant) so dividing the ergo power setting by wheel speed gives the torque at the Flow's roller.
What you will see is that the groups of points are far from horizontal so the error between the power desired and actual power (as measured with a Power Tap) is not at all constant. At high wheel speeds the Flow is unable to generate enough load to maintain constant power. Whether that inability is related to the load generator itself (which is also used in the I-Magic) or the Flow's control head, I could not determine.