You have confused the PHT transformation with the PHT layers of the block cipher's round function:
The PHT transformation maps a two-byte input to a two-byte output. It uses the formula you listed. For instance, PHT(0,1) = (1,1).
The PHT layers of the round function map a 8-byte input to a 8-byte output. Each round has a sequence of 3 layers. Each layer consists of 4 parallel PHT transformations.
The input/output vector you list is for the PHT layers of the round function, not for the PHT transformation itself. If you follow the computation diagram (see, e.g., Figure 7.12 and Section 7.7.1 in the appropriate chapter of the Handbook on Applied Cryptography, you will see how the PHT layers transform the input (0,0,0,0,0,0,0,1) to (1,1,1,1,1,1,1,1) by application of 12 PHT transformations in the appropriate sequence.
One last comment: I do not recommend using SAFER K-64. It is an outdated cipher. There are a number of shortcomings: there are some minor cryptanalytic weaknesses in it; it has not been studied very much; and worse, the key size and block size are too small. For these reasons, I do not recommend that you use SAFER K-64 or that you expose it as an option to your users. Instead, I recommend that you use AES.