That is why Johannes said “with only a bit of energy”. Although your case should nest the one without energy it’s possible for those limit cases not to work due to the issue you mention. Another example is working with a CRRA function or CES function. These nest the Log and Cobb Douglas function but you cannot just change the parameters to 1.
Try something like alpha = 0.99 if it runs as expected then you know the root of the problem is this new sector with a higher share.