log( n ) is sufficiently small and grows sufficiently slowly that other factors (e.g. memory use) had a bigger effect on the running time than the size to find the word in the dictionary.

The running time to find an element in a balanced BST is O(1) so we would not expect the running time to get larger as the dictionary got bigger.

We were measuring the best case performance of the dictionary, which does not change as the dictionary size grows.