I think that is not true, especially when taking into account that Sqrt[ 196 -x^2- y^2 ] is not a complex function. The flaw is the strict confidence in the Limit Procedure, which gave
1/4 \[Pi] (-196 + y^2)
as value of the inner integral at the upper bound. When I had a close look at this limit
Limit[j0 /. x -> uu Sqrt[14^2 - y^2], uu -> 1]
it turned out that the negative value must be used. So writing
j2 = -j12 - j11 // FullSimplify
instead of the j2 given above and
jj2[a_, e_, y_] := Evaluate[j2]
and
ff[a_, e_] := NIntegrate[ jj2[a, e, y], {y, 0, ((-a)*e + e*Sqrt[196 + 196*e^2 - a^2*e^2])/(1 + e^2)}]
which when plotting gives the same picture as the other pics above
Plot3D[ff[a, e], {a, 0.01, 2}, {e, 0.01, 2}, ColorFunction -> "SolarColors"]

What is the learning? Mathematica is really a great software, but you can't always believe everything.
The flaw was that I didn't check if this limiting process gave a correct answer.