Ok, so I found Table 10 in the manual which describes how offsets are calculated using the two bytes. That seems to corroborate my understanding described above. So you can either look at it as one byte is the integer portion and the two bits determine the fraction, or you can take all 10 bits as the value, with an LSB of 0.25 and calculate the temp that way.
HOWEVER...looking at Table 10, that all makes sense, except for the offset value of -0.25. I don't understand how a High byte of 0xFF and a low byte of 0x80 can produce -0.25. Doesn't that have to be -0.50?