For very low-power (peak power is important! Capacitors can supply some sub millisecond only; bear this in mind if the device uses a radio modem or other high current peaky loads), LDO linear regulator isn't necessarily a bad choice at all. With low output current, dropouts down to around 100mV are specified, and the open-circuit voltage of an empty cell is some 3.4V, so there is little or no wasted capacity at all. The reason why cell datasheets specify discharge cutoff like 2.5V or 2.8V is that cell DC resistance rises sharply when the cell is getting empty, causing increasing voltage drop when there is significant load current (like C/2 or C/5 as typically specified in cell datasheets). But for low current projects that need only maybe some C/20 peak current, 3.4V cutoff is acceptable and maybe loses only some 5-10% of capacity, definitely not "half".
In low power devices, efficiency is dominated by quiescent current draw anyway and it's hard to beat some 70% real-world efficiency of an LDO linear regulator with a switcher.