The Power Of Two Minor Pieces

IT is well known that a bishop and knight are, other things being equal, superior to a rook and pawn, even though the value of both pairs is 6pts by the traditional counting method (Q-9, R-5, B-3, N-3, P-1).
But consider the following position, reached after a sequence such as 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Nc3 Bc5 5.Ng5 0-0

This position occurs 741 times in ChessBase's 2020 Mega database

In 141 games - almost a fifth of the total - White played either 6.Nxf7? or 6.Bxf7+?
Clearly the superiority of bishop and knight to rook and pawn, especially in the opening, is not so well-known as might be thought.
I believe the traditional counting method is largely to blame for this ignorance.
Database computer analysis by IM - now GM - Larry Kaufman shows the true average values are more like Q-9.75, R-5, B-3.25, N-3.25, P-1.
So giving up bishop (3.25) and knight (3.25) comes to 6.5pts.
But it is worse than this because the same computer analysis shows the bishop-pair to be worth 0.5pts, so in the example above White is giving up 7pts (B-3.25, N-3.25 and bishop-pair-0.5).
A rook is usually most powerful in endings where there are open files for it to operate on, whereas a knight is much more of a middlegame piece.
A search of my personal database reveals I have 30 times had an ending of bishop and knight versus rook, with various numbers of pawns on each side.
Some idea of the power of a rook in endgames can be gathered from the fact that in those 30 games, the player with bishop and knight won only 11 times, compared with 10 wins for the player with the rook (and nine draws).
Here is a recent correspondence example.

Black to make his 62nd move in Alexander Sherwood (2320) - Spanton (2260), 5th British Webserver team championship (division two) 2018

*****
*****
*****
*****
*****
Black has an extra pawn, and connected passers; White has a farside passed pawn.
62...Rb8
62...Rd8 is met by 63.Nc6, preventing ...d4.
63.Nb5 h4 64.Kf2 Kg5 65.Ke3 Re8+ 66.Kf2
White offered a draw.
66...Kf6 67.f4 g5 68.fxg5+ Kxg5 69.Nc7 Re5
69...Rd8?? 70.Ne6+.
70.Nb5 Kf4 71.a5 d4
There is no other way to make progress.
72.Nxd4 Rxa5 73.Bg6 Ra8 74.Bf7 Ke4

White to play and draw

*****
*****
*****
*****
*****
75.Bxc4!?
This shocks - assuming analysis engines can be shocked - Komodo10 and Stockfish10, whose evaluation changes from a small edge for Black to Black is winning.
But after …
75...Kxd4
… AS successfully claimed a draw, which indeed it is according to the Nalimov endgame tablebase (but I bet he would not have played 75.Bxc4!? over the board!).