Question: Nonlinear fit of difficult function

I have a data point set:

x_val:=<250,300,350,397,451,497,547,593,647,691,745,788,840,897>:
y_val:=<0,0.5,2,6.3,23.2,48.7,71.2,83.4,90.1,92.8,94.7,95.7,96.9,97.8>:

I want to make a least square fit using this difficult function:
 

function:=x->1-exp(-(k*exp(-(E/(8.314*873.15))*((873.15/x)-1)))*(0.026/350))

but both Statistics[Fit]:
 

with(Statistics):fit_nelog:=Fit(1-exp(-(k*exp(-(E/(8.314*873.15))*((873.15/x)-1)))*(0.026/350)),<x_val|y_val>,x,parameternames=[k,E],output=[parametervector,residualsumofsquares]);

and DirectSearch[DataFit]:

with(DirectSearch):fit_nelog2:=DataFit(1-exp(-(k*exp(-(E/(8.314*873.15))*((873.15/x)-1)))*(0.026/350)),x_val,y_val,x,method=cdos);

give wrong k,E parameters. The correct parameter values were obtained with Excel Solver:

k=27843.3551042397

E=68.4

The approximately correct parameters were fitted when using logarithm form of the function.
How can I obtain correct parameter values in Maple using given form of the function?