Tolerant edges, also called "gaps" are not actual "holes" in the geometry. To be precise they are surface-to-surface intersections that exist at a lower tolerance. They also should not be considered a measure of quality. Every contemporary CAD system uses tolerant edges and creates them during modeling operations. There are many modeling operations that could not succeed without tolerant edges.
In TransMagic the "Show Tolerant Edges" function exists as simply another metric with which to analyze the model. Here's an example:
Let's say a cube was created in CATIA V4. CATIA V4 is a fairly low precision system and TransMagic is a very high precision system. In fact the value that TransMagic considers zero (0) is 10e-6 mm, which is smaller than most CAD systems. Depending on how the cube was created, TransMagic could report all 12 edges as tolerant because this measurement is relative to TransMagic's zero value. Then let's say we run Full Repair on the cube. Since the cube is comprised of 6 planar surfaces, TransMagic is able to extend and re-intersect all six surfaces perfectly due to the underlying infinite plane of each surface. So Full Repair will in this case will produce zero tolerant edges - meaning that every surface-to-surface intersection in the cube is now precise to TransMagic's zero value. However, this is not to be expected as the "correct" result as there are countless other cases where an underlying surface cannot be extended and thus tolerant edges will remain. This is perfectly acceptable and correct in all contemporary CAD systems.
Also keep in mind that since the TransMagic "zero" is very precise that what TransMagic detects as a tolerant edge may be considered precise to zero in the receiving CAD system. Therefore, tolerant edges should not be considered a measure of quality but rather a relative metric to gauge the geometric composition of a part. It's also a useful metric to gauge geometric composition before and after Full Repair.