integrate bessel_k(2,%i*y) surprise



I am just playing around with bessel_k and don't
understand the large imaginary part of the integral
of bessel_k(2, %i*y) over the interval
1 <=  y <= 100, based on the table of discrete 
values displayed below:
(f(x) is just fchop(expand(float(x))) )
--------------------------------
(%i1) load(nint);
(%o1) "c:/work2/nint.mac"

(%i25) for y:1 step 10 thru 100 do 
          print("  ",y,"  ",f(bessel_k(2,%i*y)))$

   1    0.18048997206696*%i-2.592886175491198 
   11    0.21841533174753*%i+0.3119789483129 
   21    -0.031858737423089*%i-0.27222015896945 
   31    0.20481495858656-0.093918476739506*%i 
   41    0.16377581778393*%i-0.10738879820159 
   51    0.0024786070162412-0.17554486214757*%i 
   61    0.13641169853138*%i+0.084590710242742 
   71    -0.064015208750726*%i-0.13429138505077 
   81    0.13815292462619-0.017659551649835*%i 
   91    0.084630268914206*%i-0.10051430087896 

(%i26) f(integrate(bessel_k(2,%i*y),y,1,100));

(%o26) 1.736293756153568-2.1367387806763253E+42*%i
----------------------------------------------------------------
Where does the 10^42 come from??

Ted