Assume that W is the test window size, for each testing interval, except for the putative QTL, two markers that are at least W cM away from the testing interval (one for each direction) are first picked up to fit in the model to define a testing window for blocking other possible linked QTL effects on the test.
