What a phenomenon! It appears that each too-much arc segment is the rest of the circle that the arc segment you want is a part of. The bulge factors have gotten screwed up, both in reversed sign and also numerical value, though what would do that, I can't imagine.
If no catch-all solution is found, but if, in fact, each is the rest of the circle that the arc segment you want is a part of, and therefore is of the same radius and shares a common center point with it, you can fix them with FilletMatchRad.lsp and its FMR command, available >here<. It asks you to select a curve to get a radius from -- pick the "wrong" most-of-a-circle arc segment. Then pick the adjacent line segments, and it will Fillet them using the same radius, which presumably will correct the situation. If you know that any of them are of the same radius, you don't need to start over for each one, but can pick just one "wrong" segment to get that, and then pick as many pairs of straddling line segments as you want.
Kent Cooper, AIA