I liked the suggestion "the marked assumption," found in the comments. Unlike Willie Wong's suggestion the "contradictive assumption," the "marked assumption" only partially suggests the nature of the expression. This is hard to fix unless the expression is made standard.
An advantage of the "marked assumption" is that it can serve double-duty: Even in cases where the truth of the assumption is unknown, the expression is still relevant. This is particularly notable in the study of the Riemann Hypothesis where its truth is occasionally assumed and consequences derived. If the marked assumption (the Riemann Hypothesis) holds, then all we have done is discover consistent structure. If the marked assumption fails, then our presumptive work can just be seen as a search for a contradiction.