When batch analysis is run against an uncalibrated channel, it doesn't attempt to fit GMMs, but returns NaN for them instead.