When i wrote an MBR/VBR for some real mode x86 stuff, i initially relied but on the wikipedia and osdev pages for FAT and the bios/bootlader stuff. This is how i learned the hard way how unreliable it is. Details are left out, the scope of validity for information is rarely specified, vague "many implementations" phrases that do not give any hints whether a certain behavior needs to be emulated or not.
I ended up discarding both wikipedia and the osdev wiki and using the IBM PS/2 reference manual, as well as some details from the MS PnP specification.
I had similar issues with the wikipedia explanation of the german Hl- and Ks-Signal systems for train safety, as well as articles related to intel peripheral controllers and serial ports. Often they are just not deep and detailed enough to answer my questions.
[tl;dr] It feels like reading a pop-sci article when you actually need the original paper.