In this paper, we consider a two-hop cooperative network with a direct link based on an energy harvesting (EH) decode-and-forward relay. The energy-buffer equipped relay harvests energy from the ambience, and uses the harvest-store-use (HSU) architecture. Since it is known that using a discrete-state Markov chain to model the energy buffer is inaccurate even for moderate number of states, we use a discrete-time continuous-state space Markov chain instead. We derive the limiting distribution of energy for both the incremental on-off policy (IOFP) and the incremental best-effort policy (IBEP), and use them to obtain expressions for outage probability and throughput. The corresponding expressions for non-incremental signaling follow as a special case. We show that the stable buffers using IBEP harness a diversity of two as compared with those using IOFP, which attain a diversity of one. However, while buffers using IBEP are consequently more reliable than those with IOFP, their throughput performance is only marginally superior. The simulation results are presented to validate the derived analytical expressions.