The "one forecasted path" you see generally speaking is a blend of roughly 15 or so forecasting models.
Some models may nail the path, some won't. Plus, as each "sample of weather data" is taken, a new forecast is issued, once again a blend of the multitude of models.
The data used to forecast is ever changing as the weather evolves, therefore if one piece of the puzzle isn't the same as what it was forecasted to be earlier, every model therefore may have an individual change, thereby changing the blended "single path" forecast.
Keep in mind, they ballparked this one pretty damn well. A week ago, the forecast was it's gonna come in somewhere in southern Florida...which it has. Yes, it's edged westward as we got closer, but closer generally lends to more accurate short term forecasts versus the long term forecasts of a week ago.