GJR-GARCH modelling in EViews

I am trying to use a GJR-GARCH model for returns of ftse100 index over the last 20 years using daily data.
I have run: "rftse c"
However, in the variance equation, the ARCH parameter (alpha) has a negative coefficient, which, as I understand, violates the non-negativity conditions. As in Brooks book: intercept and arch term must be >0. And the Garch term (Beta) must be greater than or equal to zero. And alpha + gamma (leverage term) must be greater than or equal to zero.
I have tried using regressions using AR and MA terms eg. "rftse c ar(1) ar(2) ma(1)", but the alpha term coefficient is always negative.
With other indices I have not had this problem.

Any help is appreciated,