The first plot is the (normalized) raw color-matching data from which the CIE 1931 XYZ space is derived. They have been normalized to have equal areas under the curves. If you multiply the green by 1.3751 and the red by 72.0962, you will recover the actual intensities of the red, green, and blue primary monochromatic sources (at wavelengths 700., 546.1 and 435.8 nm) required to match a unit intensity monochromatic source at wavelength λ. Notice that at those primary wavelengths, two of the curves are at zero, as you would expect. The red primary had to be at a high intensity since the sensitivity of the eye is so low at that wavelength (700 nm).
The second plot is the tricromatic coefficients. If the first plot is of Wi(λ), where i is R, G, or B, then the second plot is a plot of Wi(λ)/ΣiWi(λ). As you would expect, at the primary wavelengths, one curve is at unity, the other two at zero, and the sum of all the curves is unity at every wavelength.
For the third plot, the raw data was re-measured using different primaries at 650, 530 and 460 nm. The trichromatic coefficients were re-calculated from this data, and are shown in the third plot. This is from the paper "A re-determination of the trichromatic coefficients of the spectral colours" by Wright (1929).