Our team has applied the dissolve tool to reduce the number of records in a linearly referenced highway network for a state. We have noticed that perhaps one percent of the local roads, mostly dead ends are being dropped. We've learned this so far: in some but not all cases, the dropped records have bad (milepost) measures and when we replicate the problem starting from a subset of the LRS, the dissolve tool produces fewer dropped records than for the same location when the dissolve is applied on the entire network. The base/original LRS contains about 2M records. Can anyone offer some insight on what could be causing the problem and how to address it?