girder | A data management platform for the web , developed by Kitware | Data Visualization library

Try this:

ggplot will try to combine legends wherever possible by default; however, you will override this combination if any part of the legend creation, labelling, or guides are differentiated for the two aesthetic modifiers in question. In this case, you have shape and color, which are combined by default. The problem is with this line:

You can make a group based on both UG and S variables using a composite variable UGS:

Here's your output using the sample data:

There are two issues. First, legend.margin has negative parameters on the top and bottom, leading the box to be smaller than its contents. You probably want positive numbers for each. I picked legend.margin = margin(2,5,2,4, unit="mm") below, to add 2mm padding on top and bottom. (The parameters are in order of top, right, bottom, left.). [EDIT - I had accidentally shown wrong order]

Second, you have white sub-legend edges that are overlapping on your overall legend, creating uneven line thicknesses.

We can see this if we make the sub-legend outlines red:

From your output example I suppose that you want group your entities by their zone property, and get result as an array.

The code that do it:

As already suggested by @StupidWolf in his comment, you main issue is that you mapped Sample on color instead of EQ. And there is no need for separate geom_point layers. One geom_point is sufficient:

Instead of using str_replace I guess it would be easier to extract what you want using regex.

The issue is that you are overriding the color in the legend via your geom_segments when settinng color="black". To prevent this you could add inherit.aes=FALSE to each of your geom_segements or set the aesthetics inside geom_line.

BTW: You have two scale_color_s in your code so I dropped the scale_color_hue

Using mtcars as example data:

You should use a continuous scale and specify the axis breaks that you want: