Traditionally, how to reduce energy consumption has been an issue of utmost importance in wireless sensor networks. Recently, radio frequency (RF) energy harvesting technologies, which scavenge the ambient RF waves, provided us with a new paradigm for such networks. Without replacement or recharge of batteries, an RF energy harvesting wireless sensor network may live an eternal life. Against theoretical expectations, however, energy is scarce in practice and, consequently, structural naiveté has to be within a MAC scheme that supports a sensor node to deliver its data to a sink node. Our practical choice for the MAC scheme is a basic one, rooted in ALOHA, in which a sensor node simply repeats harvesting energy, backing off for a while and transmitting a packet. The basic medium access control (MAC) scheme is not able to perfectly prevent a collision of packets, which in turn deteriorates the throughput. Thus, we derive an exact expression of the throughput that the basic MAC scheme can attain. In various case studies, we then look for a way to enhance the throughput. Using the throughput formula, we reveal that an optimal back-off time, which maximizes the total throughput, is not characterized by the distribution but only by the mean value when the harvest times are deterministic. Also, we confirm that taking proper back-off times is able to improve the throughput even when the harvest times are random. Furthermore, we show that shaping the back-off time so that its variance is increased while its mean remains unchanged can help ameliorate the throughput that the basic MAC scheme is able to achieve.