the code example provided

No!

The first one is correct, because it is adding an offset (in bytes) to the starting address. This is the type of thing you would do when receiving the data from a byte stream.

The second one is adding an offset in dataLength type sizes, which would be an error in this case.

Example:

int dataLength; // assume 32-bits (4-bytes)
size_t cbLeftToReceive = 2; // count bytes left

// OK: Equivalent to (BYTE*)(&dataLength)[2]
BYTE* bp = (BYTE*)(&dataLength) + sizeof(dataLength) - cbLeftToReceive;

// ERROR: Equivalent to (BYTE*)(&dataLength)[8]<br>
BYTE* bp = (BYTE*)(&dataLength + sizeof(dataLength) - cbLeftToReceive);</br>

er...Yes.

Keep it all in one chunk.

The both forms are intuitive incorrect,
your origin lies at the address of a variable, not at a data position... :)

Just post the whole function here :-D