RLP decoding rules accept invalid encoding [129, 1]?

If I'm reading the rules under 'Decoding' at the end of data.md correctly, it looks like invalid encodings like [129,1] (where I'm informally writing strings as lists of codes for simplicity) are accepted by the rules:

rlpDecode([129,1])
==>
rlpDecode([129,1], decodeLengthPrefix([129,1], 0)
==>
rlpDecode([129,1], decodeLengthPrefix([129,1], 0, ordChar(substrString([129,1], 0, 1)))
==>
rlpDecode([129,1], decodeLengthPrefix([129,1], 0, 129)
==>
rlpDecode([129,1], str(1, 1))
==>
substrString([129,1], 1, 2)
==>
[1]

Please correct me if I'm misreading/misinterpreting the rules somehow.

Note that [129,1] is invalid because the singleton string [1] should be encoded as itself. For instance, see the test "byteShouldBeSingleByte01" at https://github.com/ethereum/tests/blob/5d85b07e460e38a991801087f967d308346b32ab/RLPTests/invalidRLPTest.json#L37.

Start with the Decoding rules at the end of data.md and compare their behavior with the byteShouldBeSingleByte01 case in the linked invalidRLPTest.json. Confirm whether [129,1] is accepted, then update the rules so the invalid encoding is rejected while the canonical singleton encoding remains valid.