I find that charting the room can help with figuring out eyeline. Here’s the overhead of your scene.
The two characters are the dots matching their shirt colors. Green camera 0 is the wider shot. The yellow line is the eyeline created between them that indicates the 180 degree rule. With your establishing shot, you need to stay out of the yellow dashed area. You can go there but it’s best to avoid it unless you’re doing it intentionally. The two cameras labeled 1 and 2 represent the two shots you’re considering.
You can confirm this by looking at the direction of the arrow from the point of view of each camera. From angle 1, the arrow points right; from angle 2, it points left. From angle 0, it points right, so to stay on the same side of the line, you need red to be looking right.
You can figure this out without a chart, too. In your wider shot you’re seeing red’s right cheek. In general, that’s the cheek you should keep seeing. If you see the left cheek, you may have crossed the line (unless the eyeline changed).
So, shot two would cross the line. In order to see red’s left cheek you’d have to be looking over blue’s other shoulder if they’re meeting each other’s eyes.
I have a guess why it might seem better to go with angle 2, or at least why angle 1 might seem wrong. I’ve added another arrow to the image above:
This short arrow is how the body position of red appears in the scene. Red seems to have his chest open to blue’s left rather than being the same as his eyeline. You can see red’s right shoulder closer to camera. In angle 1, it seems to be farther back than the left shoulder, making it seem like you’ve crossed the line of his body even though you haven’t crossed the eyeline.
So, you’ve mixed angle 1 so it can seem a bit odd. In angle 2 the eyeline and body position both point left so even though you’re crossing the line, the body and head are aligned. In fact, the body position in both angles 1 and 2 look identical, which seems wrong.
If you square up red’s shoulders with the eyeline, I think it might improve the shot.