These charts are flow cytometry graphs, something not explicitly stated in the figure description, but obvious to everyone who has seen similar graphs before (and I assume it's at least mentioned in the method section of the paper, I didn't read that though).
Each dot in a flow cytometry graph represents a single cell and, as you correctly identified, the PE & FITC flourescence corresponds to fentanyl and the respective proteins of interest. (The labelling was done somewhat indirectly and as is typical for flow cytometry with antibodies but that doesn't matter for the interpretation.). The dots are coloured according to their density: blue indicates only single dots and warmer colors show increased number of overlapping dots (with red being the highest).
It is also important to note that all the axes in these graphs have logarithmic scales, so the shifts you see in the graph represent shifts in orders of magnitude.
The combination of the logarithmic scale with per cell data points means, that you also see 'negative cells' on each axis: along the FITC axis (which represents expression of the protein) you can make out two groups of points. The lower one of these groups has cells that don't express the protein (meaning the florescence detected is just background from staining), while the higher groups contains cells that do express the protein (the staining for the protein is significantly above background).
If you now compare the two proteins you can see that high (protein expressing) group for Fen49 also has an increased signal for PE (which represents fentanyl), meaning that these cells bind fentanyl. In contrast with that for 2QZ3 the fentanyl signal is independent of whether the cells have a low signal or high signal for protein expression.
The bar graph on the right gives the average PE (fentanyl) levels for other additional protein variants. The figure description doesn't explicitly say whether this is the average of all cells or only the ones expressing the corresponding protein at above background levels.
