Hoppy is right IMHO. I'm accustomed to 3+3T and almost invariably the results after even longish sectors came in way under that on the B747 classic to which I was then assigned.
Apples and oranges maybe?
The IRU drift rate isn't constant. The IRU will potentially have high drift rates during the taxi out / takeoff / departure climb and also on the way down during the descent, approach, landing and taxi back to the gate.
Conversely, the IRU should be relatively stable during cruise.
On trans-oceanic flights with long stable cruise times, 3 + 3T is a good rule of thumb for approximating the drift limit. But 3 + 3T isn't a good approximation for very short flights.
And during an active approach, especially with a go-around or missed approach, the IRU can drift at much higher rates than 3 + 3T, generally assumed to be 2 nm per 15 minutes (which is 8 nm per hour) based on empirical testing.
For RNAV GPS approaches, if the GPS is lost then (typically) the system is not certified to fall back to DME/DME updating and will only rely on the IRUs.
So for safety reasons, during procedure design in case of GPS loss the IRU is assumed to be exhibiting high drift (8 nm/hr) for N minutes of activity before stabilizing back to 2 nm/hr (with N being a variable depending on what is being designed.)