im2bw get different binary results on same data between double and uint8 in R2016a

Hi, I am using im2bw to convert a RGB image to binary, but get very different results when image was in uint8 or double format.

How does this difference appear?

which results should I trust?

Thanks.

a= [];
a(:,:,1) = [ 1 2 6 7 6;
    1 1 8 7 8;
    1 1 9 6 7;
    1 2 7 8 6;
    1 2 6 8 7];
a(:,:,2) = [  0 0 0 0 0;
    0 0 0 0 0;
    0 0 0 0 0;
    0 0 0 0 0;
    0 0 0 0 0];
a(:,:,3) =      [0 0 0 0 0;
    0 0 0 0 0;
    0 0 0 0 0;
    0 0 0 0 0;
    0 0 0 0 0];

graythresh(a)%0.4980
imshow(im2bw(a)) % shows a normal binary image

graythresh(uint8(a)) %0.0137
imshow(im2bw(uint8(a))) % shows only black. All information is lost.