If you create a DTM for a block that has several image pairs and choose the single option as the output form, the points in the same ground DTM cell from different image pairs will be processed and merged together. The merging algorithm aims to choose the best point, but not average the elevations. It works in the following way: In each ground cell defined by the output DTM cell size, if there is only one point matched, this point will be kept. If there are two points matched, the one with the better quality index will be kept. If both have the same quality, the one close to the cell center will be used. In cases with more than two points matched in the cell, the median elevation will be used, since the median elevation is least likely to be a gross error point.