The Fibonacci series is just one example of an exponential estimation scale. The reason an exponential scale is used comes from Information Theory.
The information that we obtain out of estimation grows much slower than the precision of estimation. In fact it grows as a logarithmic function. This is the reason for the higher uncertainty for larger items.
Determining the most optimal base of the exponential scale (normalization) is difficult in practise. The base corresponding to the Fibonacci scale may or may not be optimal.
Here is a more detailed explanation of the mathematical justification: http://www.yakyma.com/2012/05/why-progressive-estimation-scale-is-so.html
