In a typical MOSFET, the source contact is connected to the bulk substrate. Therefore, there is a diode junction between drain and source/bulk - the typical MOSFET reverse diode. This diode's junction capacitance is the MOSFET's drain-source off-state capacitance. The drain and source potentials are "facing" each other via this reverse-biased junction, forming a plate capacitor.
In a JFET, however, there is no such reverse diode. A JFET is a true three-terminal device; the source and drain contacts are connected directly to the substrate, which is the (normally-on) channel. When the JFET turns off, the conducting path between drain and source gets pinched off by the expanding depletion region from the gate. There is no point in the JFET anymore where the drain and source potentials face each other - the gate potential "pushed" itself between them via a depletion region. As a result, there's a drain-gate and a gate-source capacitance, but no significant drain-source capacitance anymore since the gate potential is in the way. Additionally, depending on gate voltage, the width of the depletion region varies, which means that the drain-source capacitance is both negligibly small and poorly defined.
The drain-gate and gate-source capacitances also vary greatly depending on the voltage between gate and drain / source since they are caused by reverse-biased diode junctions as well.