Gene expression from both parental alleles is beneficial by masking the effects of deleterious recessive mutations and by reducing the noise in gene expression in diploid organisms. However, a class of genes are expressed preferentially or strictly from a single allele. The selective advantage of avoiding biallelic expression is clear for allelic-excluded antigen receptor and odorant receptor genes, genes undergoing X-chromosome inactivation in females and parental genomic imprinted genes. In contrast, there is no clear biological rationale for the predominant and stochastic monoallelic expression of cytokine genes in the immune system, and the underlying mechanism is elusive and controversial. A clarification of the mechanism of predominant monoallelic expression would be instrumental in better understanding its eventual biological functional. This prompted the development of a quantitative framework that could describe the dynamics of the pattern of allele expression of the IL-10 gene, from which general quantitative insights could be gained. We report that the experimental observations on these patterns of allelic expression cannot be easily reconciled with a simple model of stochastic transcriptional activation, in which the two alleles are, at any time, equally competent for transcription. Instead, these observations call into action a general model of eukaryotic transcriptional regulation according to which the locus competence for transcription is dynamic, involving multiple, cooperative and stochastic modification steps. In this model, the probability that an allele becomes transcriptionally active is a function of the number of chromatin modifications that it accumulated. On the basis of the properties of this model, we argue that predominant monoallelic expression might have had no adaptive role, and may have evolved under indirect selection for low frequency of expressing cells.