Sequence comparison is a widely used computational technique in modern molecular biology. In spite of the frequent use of sequence comparisons, the important problem of assigning statistical significance to a given degree of similarity is still outstanding. Analytical approaches to filling this gap usually make use of an approximation that neglects certain correlations in the disorder underlying the sequence comparison algorithm. Here, we use the longest common subsequence problem, a prototype sequence comparison problem, to analytically establish that this approximation does make a difference to certain sequence comparison statistics. In the course of establishing this difference we develop a method that can systematically deal with these disorder correlations.