Two independent sets of recent observations on newly sequenced microbial genomes pertain to the prevalence of short inversion as a gene order rearrangement process and to the lack of conservation of gene order within conserved gene clusters. We propose a model of inversion where the key parameter is the length of the inverted fragment. We show that there is a qualitative difference in the pattern of evolution when the inversion length is small with respect to the cluster size and when it is large. This suggests an explanation of the lack of parallel gene order in conserved clusters and raises questions about the statistical validity of putative functionally selected gene clusters if these have only been tested against inappropriate null hypotheses. Keywords inversions, conserved segments, gene clusters, evolution