CPython's Lib/heapq.py is implemented using the binary heap. Your implementation is based on another data structures with different asymptotic runtime complexity.

As such, it's totally possible there exists a graph where your implementation can perform faster than the official one. In any case, it's important to benchmark both of them on different graphs.

References:

• https://en.wikipedia.org/wiki/Binary_heap
• https://docs.python.org/3/library/heapq.html