Ideal air standard cycle efficiencies of very conventional ICEs can reach 60-70+ percent, and this is likely the figure that peddlers of newfangled engines quote half-truthfully, thinking that the people who matter (investors for money and the public for mindshare) are ignorant enough to buy it (and it appears to work often enough).
However, the air standard efficiency is NEVER reached in a real engine, nor even closely approached. Inventors that devise new engine ideas that still run on some variant of established thermodynamic cycles think they can return to or exceed the ideal efficiencies, but they're flat wrong. For automotive Diesel engines, a best point of 43% BTE has been reached in production. For large engines, it can be appreciably higher, culminating in 55+ percent for marine engines. Research has been taking place to bring efficiencies in on-highway trucks to 60% BTE by carefully identifying and addressing all sources of losses, and employing some sort of exhaust energy recovery system like turbocompounding, a bottoming cycle or thermionic converters.
In contrast, many like to malign the "wastefulness" of ICEs, and wax lyrical about fuel cells and EVs. Well, in a PEM fuel cell, the ideal open-circuit cell efficiency can be over 90%, but again, this is NEVER attained nor approached by a long shot. After accounting for all the sources of losses, the electrical output of the fuel cell stack yields an efficiency of about 50%; values as high as 65% have been attained in labs and is dependent on conditions, but this is not the end of losses. Some power is use to drive an air blower or compressor that is used in a PEM-FC to increase the power density, then there losses in the power electronics and electric motors (the latter between 80-90% efficient over a wide operating range). These losses are fairly fixed, and at very low demanded load, the system efficiency is accordingly very low to offset these parasitic losses, even though a FC is theoretically most efficient at low current loads. All this doesn't even consider the energy conversion efficiencies of the hydrogen fuel from a primary energy source that fuel cells need (either generated somewhere or reformed onboard).
Ditto EVs. Many people claim efficiencies of 80-90%, but consider only the losses at the electric motor. There are also losses through the power electronics (~90% efficient), as well as charge- and discharge losses (~80% each, mostly depending on battery type). Considering only these losses, the system efficiency drops down to something more like 45-55%. Hmmm... not so far off from the best ICEs. Again, the primary energy source has to some from somewhere. If it's generated from an ICE acting as a range extender, then your limiting factor is actually the efficiency of the ICE(!); if the electrical power comes from the grid, the efficiency depends on the mix of generating sources, but an efficiency figure of 40%-50% is typical for a wide range thermal power plants that include steam- and gas turbines using coal, oil, natural gas or nuclear as the primary energy source.
So let's recap. We've seen that regardless of the technology used, we come to a total efficiency wherein still a big chunk if not a majority of available energy is actually lost. Nobody has credibly come up with (nor probably will in my lifetime), any solution so revolutionary that will raise total, real-world efficiencies far above where they are today. And that's not the problem: as "wasteful" as all these current solutions are, the bigger problem is that transport applications are rarely operating anywhere near the potential best efficiency points of ANY of the abovementioned solutions. Regardless of ICE, FC or EVs, if these "wasteful" devices could only operate close to their (still terrible) peak efficiency most of the time, the real-world fuel consumption -- and your fuel bill -- could be reduced by as much as half. And the thing is, there is ABSOLUTELY no magic in it.
That's what hybrids are trying to do: trying to get the wasteful ICE to operate more frequently at operating points that are more efficient, and less or not all all when it isn't. It really is all it comes down to.