It is justified. It's just that the justification uses an extended definition of convergence. No one is arguing that the limit of the partial sums in the standard sense is -1/12, just like no one argued that there is a real number whose square is -1. Instead, there is a new kind of smoothed convergence and a new kind of complex number. These extended tools agree with all of the old calculations, and give new results for calculations that weren't possible before.
They aren't the same. An analytic continuation of a function is not the same as the function itself. The gamma function is not just an analytic continuation of the factorial function- it is another function that has the factorial as a subset of itself.
That the Riemann zeta function has a higher order function that includes it as a subset is not a surprise. However, it is a subset of the higher order function, the higher order function is not just a continuation of it. The fact that people arrived at the analytic "continuation" of the function second... well, that means nothing. It's not a continuation- the "analytic continuation" is a higher order function that uses a specific value of another function to seed itself.
Fun with Grandi's series (it is all of the following):
(1+1+1+1...) - (1+1+1+1....)
1-2+3-4+5-6...
(1+2+3+4...) - (1+2+3+4...).
(.1 +.1 +.1 +.1...) - (.1+.1+.1+.1...)
To get any number you want, just change the order of terms around, since it's all
indeterminate form to me.