I delved in my sources, and it seems that microchip uses a different algo, I added a CCIT implementation in the same vein for reference.
I have the references for your implementation now and will include it.
It is called a CLASSB_CRC16 (without reflection) a.k.a. RF69 as used in e.g. radio communications and sensors. See ATMEL, Microchip, ISO/IEC CD 14443-3.3 appendix B.
It is NOT CCITT compliant, although it uses the same start value of zero, $1021 it uses a different polynomial using only 15 bit, but a very common implementation in hardware, so it should be in.