The use of search algorithms requires the definition of a fitness function that guides the algorithms to find an optimal solution. The definition of a fitness function may require the use of a normalization function for various purposes such as assigning equal importance to various factors constituting a fitness function and normalizing only one factor of a fitness function to give it less/more importance than the others. In our previous work, we defined various branch distance functions (a commonly used heuristic in the literature at the code-level) corresponding to the constructs defined in the Object Constraint Language (OCL) to solve OCL constraints to generate test data for supporting automated Model-Based Testing (MBT). The definition of several of these distance functions required the use of a normalization function. In this paper, we extend the empirical evaluation reported in one of the works in the literature that compares the impact of using various normalization functions for calculating branch distances at the code-level on the performance of search algorithms.The empirical evaluation reported in this paper assesses the impact of the commonly used normalization functions for the branch distance calculation of OCL constraints at the model-level. Results show that for one of the newly studied algorithms Harmony Search (HS) and Random Search (RS), the use of the normalization functions has no impact on the performance of the search. However, HS achieved 100% success rates for all the problems, where RS obtained very poor success rates (less than 38%). Based on the results, we conclude that the randomness in creating a solution in search algorithms may mask the impact of using a normalization function.