Basically, a 5-stop filter factor for an R72 means that you're assuming that the amount of IR "seen" by the film---considering both the filter's curve and the film's response---is 1/32 the intensity of visible light seen by the meter. Well, as you change films, the amount of visible light doesn't change, but the amount of IR the film sees will: for instance, HIE sees much more, because it responds to high-IR at wavelengths way, way past the cutoffs of the other films.
Rollei IR400 is the opposite; it's a pretty fast film, but it's sensitive to IR only up to a fairly low level---not much above the point where an R72 filter really starts transmitting, so when you put the film and the filter together you limit the "effective" light to a very narrow band, so it takes a lot of compensation.
You can usually get away with *less* compensation at morning and evening than at midday, which is kind of counterintuitive. The same effect that causes sunrises and sunsets to be red means that there's more IR around then too, relative to visible light---the total light is less than it would be at midday, but the portion of it that's in the infrared range is relatively high, so you don't have to compensate as much from a visible-light meter reading.
Results vary with latitude and weather, too. Experiment, bracket, and consider using a compensating developer.