Ok, i try to explain it by a simple example based (fake data from my short journey just to show what happens)
Assume a user who has seen 3 pages, 2 of them twice. you can see what pages the user has seen by the "hit depth column", means you can follow the path through the pages (red line).
now lets look at the "Exits" metric. According to the description Exits the metric "number of times a given value is captured as the last value in a visit". in this visit the last value captured for page is "....cs group n" (first item), thus you get a value of "1" for the "Exits" in the first row (blue box)
now lets add a hit container and see what happens. I apply a hit segment to only show hits for page 2. I would assume that I wil only get pages views for hit 3/4 while all other hits disappear. but what happens to the exits?
Ok, that is interesting. you can see the following in the new table
as expected, the page views are now filtered to the page where I set the hit segment. that means I only get "Page Views" for the corresponding pages (green box, second row)
the "Exits" is still at the same place (blue box) and pointing to the same page as in the first table! that means, even if I do not have the first page in my segment (no "Page Views", see first column) the row appears and indicates a "1" for the "Exits". This is because it is still the same visit and this visit has seen the page in the first row as last dimension item. Therefore the page appears and gets a counter for "Exits". in the description (link above) you can read: "When applied to a dimension, Exits are counted on the last value of a visit, which may happen on any hit during the visit."
I hope that this example helps to understand why you see values outside of your hit segment.