If your results cannot be reproduced by others, and you believe your results are valid, the best next step is to look over what you published. I've seen an occasion or two where a person's description of their experimental design leaves out some step they do automatically (without thinking) but which others may not do. I think that's as rare occurrence but not to be overlooked-- it happened to me many years ago.
A very good point. I've run into this problem while trying to find bugs in my database. The end user reports a problem and tells me how they got there, but I am unable to replicate it. Then I have to go through it with them, watching over their shoulder and sure enough they forgot to mention something that they do automatically that they hadn't considered.