Continuous glucose monitoring (CGM) systems is an emerging technology, which provides detailed information on glucose patterns and trends, and alarms that alert the patient to both hyper- and hypoglycemia. However, experiments with these devices show several limitations to their optimal use for modeling of the glucose-insulin system (e.g., spurious dropouts which do not reflect the true glucose level). In order to reconstuct glucose patterns, we propose the use of Extended Fourier Transformation for nonuniform sampled data algorithm. The effectiveness of this approach is demonstrated by clinical data from type 1 diabetes patients.