Two important issues for pattern recognition in sequences are raised by this illustration and warrant discussion even if they fall outside the strict reporting of a kernel density distribution method. Firstly, it is clear that for any fixed resolution, L, all conserved segments of longer length will have its L-long sub-segments represented as peaks scattered throughout the distribution. As a consequence, the choice of value for the smooth parameter, S, should be set as to maximize the recognition of an objective quantity, such as information content. When scanning different scales, by using various values for L, the optimal value of S would also be different, as it would be dependent on the information content encoded at that scale. Secondly, the shorter sub-segments of a conserved segment of length L, will set the base height for the quadrants where the conserved L-long segment is inserted. Therefore, the availability of a density distribution kernel for the projection of sequences in a continuous space also creates the opportunity to devise de-embedding schemes that will pinpoint the location of conservation for arbitrary target resolutions.