No single histone mark is entirely necessary for regulatory function, nor is any single histone mark entirely sufficient for regulatory function. So, contrary to what some believe H3K27Ac does not equal enhancer activity there will always be enhancers without H3K27Ac and H3K27Ac marked regions that are not enhancers, even if the overlap between H3K27Ac and enhancer activity is generally high.
In a way it makes sense for there to be H3K27Ac peaks without ATAC peaks, and in fact, it is the opposite that does not make sense: H3K27Ac signal marks the presence of histones with lysine 27 acetylated. But ATAC marks regions where there are no histones. The usual explanation of this is that it is the histones either side of the accessible region that are marked. The alternative explaination of course is that ATAC peaks are not regions that are total free of nucleosomes, but rather regions that have fewer histones present, less of the time, when averaged across the cell population. Thus accessibility is not a binary yes/no question, but a continuum, with there being nothing particularly special about the threshold we choose to decide something is or isn't a peak.
We usually think of open chromatin forming through the action of pioneer transcription factors that evict the histones and therefore leave the chromatin accessible for other TFs to bind. I'm not sure its really clear what the role of H3K27Ac is in this process - is K27Ac necessary to weaken the histone-DNA association, and allow the pioneer factor to bind? Does the pioneer factor recruit histone acetyl-transferases and therefore H3K27Ac is the consequence of pioneer factor binding? Is it both, in some sort of feedback loop? Probably all three of these possibilities are true in different cases.
This is before we even talk about the differing sensitives of your peak finding assays and whether your peak sets/signal levels are absolutely accurate measures of the truth, or noisy proxies that only correlate with what is really happening.
