Load-dump
... is a killer - your TVS has to turn a huge amount of energy into heat without going pop.
ISO7637 for a 12V system has a spike peaking at up to ~90V with a rise time of 5-10ms lasting up to 400ms from a source resistance as low as 0.5ohms. That's several hundred Joules of energy in less than half a second!
Not all of that has to go into the suppressor - only the excess above the clamping voltage (but still ~60V in your case)
On the bright side, load-dumps are pretty rare, so if it's a one-off and you don't mind the small risk, you could ignore it.
Fast transient spikes
These can reach 200V when the wipers switch off for example - provide a (high-voltage-rated) capacitive route for those to ground right near the input.
Longish-term over-voltage
Automotive electronics is often specified to survive 24V for several minutes (for when a car is jump-started off a 24V truck) and 48V for up to a minute (IIRC) as sometimes 2 truck batteries are used to provide a quick boost charge to get a car moving in extremis! Your spike suppressor may pop under those conditions.
Dropouts
Battery dropouts can also be significant, there's a test in the industry which involves a series of pulses battery voltage falling to 0V - you need to have enough internal capacitance to keep your supply rails up when that happens.
Real-world requirements specification
If you want an example of how gory this can get, Ford's electromagnetic compatibility (EMC), which includes transient testing, is available on the web:
Component EMC Specifications EMC-CS-2009
Search through it for "transient" and "dropout" to see what series-production designs are supposed to live up to!
I was rather curious about why many buck converters and linear regulators were rated up to 42V, or close to it. It's a rather strange voltage to stop at, after all - why not 40V or 45V?
It's because of something called safety extra-low voltage (SELV) -- certain regulatory standards allow voltages up to 60V under all conditions, and other ones allow up to 50V. (Unfortunately I can't find any references to standards, and even if I did, you wouldn't be able to view them for free. :-( ).
When you consider a 42V nominal automobile battery under worst-case conditions (load suddenly going away during charging), it can be higher than 42, so there's enough margin to stay below 50V and be able to pass those regulatory standards.
Above the ELV/SELV limits, you can operate but you have to use different connectors, make sure there's no exposed parts that are energized at that voltage, etc. -- so there's a quantum jump of expense and difficulty that starts at that point.
Best Answer
I think the linear stabilizer will be acceptable solution, if your gadget may eat lower voltage than 12V (about 10% lower). If yes, then you may use LM-338 for example. This stability margin (10%) is needed because the every linear stabilizer has its own voltage drop near 1V or higher. And, of cause, follow the recommendations regarding capacitors - their capacity must be enough for expected load, and their voltage must be 50-60V at least to respect your remark about real voltage pikes. If your gadget is pretty simple and has some resistance to voltage overload, then you may use only LC-filter based on ferrite-ring and a combination of big capacitor and ceramic capacitor (purpose of second one is to eat the noise of generator's rectifier and ignition of brushes). If the voltage precision is a key, then your choice is a double convert 12 to ~220 and then ~220 to 12, by using a car power adapter (inverter). These adapter should be able to give you much more than 2-4 Amps.