Wells (2007) p165 asserts that 5...Nf6 was traditionally preferred over 5...e6 because
[FEN "..."]
1. e4 c6 2. Nf3 d5 3. Nc3 Bg4 4. h3 Bxf3 5. Qxf3 Nf6 6. d4 dxe4 7. Nxe4 Qxd4 8. Bd3 Nbd7!
"is an unsound pawn sacrifice" whereas
[FEN "..."]
1. e4 c6 2. Nf3 d5 3. Nc3 Bg4 4. h3 Bxf3 5. Qxf3 e6 6. d4 dxe4 7. Nxe4 Qxd4 8. Bd3
"is much more dangerous".
He goes on to state that in the latter line 8...Nd7! may overturn the old assessment, but the implication is that 6 d4 is still more dangerous against 5...e6 than against 5...Nf6.
Regardless of the current status of this line, what was the traditional explanation for why 8 Bd3 is more dangerous in the 5...e6 line than the 5...Nf6 line? Wells does not explain this.
Wells, Peter (2007). Grandmaster Secrets: The Caro-Kann. Gambit Books.