Accuracy of sampled values in Lab tiffs

Hi in the previous post, I asked about the values of L that the sample dropper picks up from images. The image was 8 bit (non-dithered) and the value of L picked up was 38.15 rather than a multiple of 100/255 which I expected, the nearest of which is 38.04.

As it turns out, converting the image to 16bit in photoshop, the Lab value is 12337 which translates to an 8 bit value of 37.64 which is ones tep lower and even further away from the 38.15 ColorThink value.

Creating an 8 bit image in photoshop with L=38 actually creates one with L of 38.04. This when sampled also shows 38.15 in ColorThink so two separate 8 bit values seem to be being made equal by ColorThink.

I tried looking at the 16 bit images in ColorThink but it won’t do that for Lab images - is that a known bug?