get-text-property signals an error if the position argument is outside a
narrowed region. That is very inconvenient. It would be much better if
it behaved like get-char-property which does not do this.
I just tried (get-char-property 1 'face) in my Rmail buffer and got an
error, so I think the two functions are alike in this regard.
I think that an error is the correct result in that case.
(buffer-substring also gives an error when the positions are out of range.)
Why do you think otherwise?
Is there a case in which get-char-property fails to give this error?