Discontinuity in Threshold regression

[honeysito]

I wrote the following threshold program in Eviews, However I am encountering a problem of discontinuity where the SSR goes to zero in some estimations. Can anyone help?

Thanks. Here is the program

delete debt
pool debt
debt.add _ant _bah _dom _gre _guy _jam _stk _stl _stv _sur _tt
debt.genr dum?=0
debt.genr unit?=1
vector(900) ssrvec
for !j=22 to 110 step 0.1
debt.genr threshold?=!j
debt.genr dummy?=@recode(debtgdp?>!j, dum?,1)
debt.genr gama1?=(unit?-dummy?)*(1/debtgdp?-1/threshold?)
debt.genr gama2?=dummy?*(1/debtgdp?-1/threshold?)
debt.ls(n) gr? gama1? gama2? lypcini? op? infl? invgdp? gegdp?
ssrvec(!j*1)=debt.@ssr
next
ssrvec.line