Dynamic defect models are used to estimate the number of defects in a software project, predict the release date and required effort of maintenance, and measure the progress and quality of development. The literature suggests that defects projection over time follows a Rayleigh distribution. In this paper, data concerning defects are collected from several software projects and products. Data projection showed that the previous assumption of the Rayleigh distribution is not valid for current projects which are much more complex. Empirical data collected showed that defect distribution in even simpler software projects cannot be represented by the Rayleigh curves due to the adoption of several types of testing on different phases in the project lifecycle. The findings of this paper enhance the well known Puntam's defect model and propose new performance criteria to support the changes occurred during the project. Results of fitting and predicting the collected data show the superiority of the new enhanced defect model over the original defect model.