The kWh values are based on the (rated/ideal) range difference per drive. E.g. to get the consumption per 100 km/mi:
rated_range_diff * rated_consumption / distance * 100
As the range is often reduced in cold weather at the beginning of a trip, the 'range_difference' gets distorted, because as the battery gets warmer, we seem to get more range. On many of my cold weather drives, the efficiency was significantly positive, which is only because the 'range' difference was smaller than it should have been.
As I see it, it makes no sense to display the efficiency figures under these conditions. We might as well display random values. However, as it is currently implemented, the values are still displayed for some drives. For example, if there is no more range reduction at the end of a drive. In this case, the drive was probably long enough to warm up the battery, making the difference less significant overall.