Generally speaking, the more intense the tropical cyclone, the greater the depth of its core cyclonic circulation (from the surface to above 200mb in some of the most intense cases), so as intensity increases we look at an average of the winds over a greater vertical extent of the atmosphere when attempting to define the "steering currents." For example, check out the layer mean wind page here at UW:

http://cimss.ssec.wisc.edu/tropic/real-time/atlantic/winds/winds-dlm.html

You'll notice that the steering indicated for a weak cyclone (MSLP greater than 1000mb) is much different than for an intense one (MSLP less than 940mb). This process of reducing the atmosphere to a 2-dimensional flow-field is one of the reasons that barotropic models are still used for track guidance despite their great simplicity -- all in all, they work quite well, especially when the flow regime is well-defined and well-behaved as it usually is in the deep tropics. In the subtropics, things can get a little wacky as in the present case when a TC finds itself trapped in a very weak steering environment. Initial position and convective assymetries can have a part to play in determining the ultimate track in situations like this (and who can really predict where an individual convective cell is going to develop?). It's one of the reasons the global models do so much flip-flopping. When there is greater run-to-run continuity, that's usually a sign that the flow is evolving in a way that makes the track forecast less chaotic (and hopefully!) more reliable.

hope this helps,
Will