Is this a Stockfish Inaccuracy? Suggesting -16.3 before -#17!

Board setup is: white king - G1, black rook - A2, black pawn - G3, black king - E3, white pawn - D3, black pawn - G4, white pawn - C4, white pawn - B4, black pawn - D5, white knight - E6. With black to move, stockfish suggests pawn D5 -> D4 (-16.3 Point Favor), when an alternate (nonsuggested) move of pawn D5 takes C4 results in a garunteed #17 move checkmate. Is this a possible quirk of the AI saving time on a Blitz game? or should the result be to suggest checkmate instead? I bring this up to better stockfish, but I'm just a novice chess player so I'm probably wrong.

is this correct?
lichess.org/editor/8/8/4N3/3p4/1PP3p1/3Pk1p1/r7/6K1_w_KQkq_-
playing as black right?
I don't see why you wouldn't just put the white king in check again with the rook? there is only 1 square he can go to- G2, then trap him with a pawn, while your king kills the white pawns?. will have to end with your rook killing the last pawn, as rook vs night and get your king back near the white king and checkmate?

the night would have to kill all of your pawns alone or white loses. this would take 7 moves minimum, but you can avoid trapping the king and keep putting him in check with your rook and delay this

the reason I think the pawn is moved is because its the fastest things to target for the night, and moving it next to your king would mean the night would have to be sacrificed, so instead white moves a pawn, allowing the king to grab 1 white pawn and then another as whites best option is to queen, buts its not fast enough and rook will take. once black queens, knight will be forced to defend and be killed by rook leaving white king along to be mated