Spectral clustering is motivated by the extraction of arbitrary cluster shapes in numerical data. This property enhanced its popularity, and despite its theoretical base having been established for more than a decade now, variants have still been introduced until recently. This algorithm basically transforms data to a latent space where arbitrary cluster shapes become easy to process by a clustering algorithm such as k-means. However distributions actually observed in this latent space have received little attention, and many authors assume theory predictions are verified. Alternatively, this paper follows a qualitative approach to check if the ideal, theoretical structure is indeed obtained in practice. A theoretical state-of-the-art summary serves the identification of parameters commanding at the transformation. Observations drawn from our experiments lead to the identification of effective parameter combinations with associated conditions.