Yeah, I have to agree. My github account is pretty barren, mostly because the lion share of my work is done professionally on private repos simply sitting on a server behind a firewall. In fact, you'd be amazed at how much code in the wild simply sits in backed up files in someone's ~/src folder. Lots of engineers, especially older ones, never learned that kind of practice.
A big chunk of my code is C, C++, VHDL, Verilog, miscellaneous PLC/PAC dialects, etc... in other words, industrial stuff that doesn't get a lot of attention from the 'fun' startup sector. Github's numbers are most likely misrepresenting (not representing?) a HUGE section of 'hard' engineering. Add to that what you've mentioned, i.e. the copious amounts of legacy and domain specific stuff still lying around, and it means you've got a hugely biased sampling.