I don't think there's anything wrong with the direction of the diodes. The circuit looks to me as some variation of what is called a Null-and-Balance detector for FM. But in this type of detector which is again a modified Foster-Seeley, the input comes via a split capacitor section (in parallel with the secondary winding) rather than a split secondary winding. This is done to improve symmetry around the center frequency, which is difficult to with a split inductor where you need exact equal inductances. The parallel load resistors over the diodes is characteristic of this type of detector when used with the split capacitor input.
On the page **broken link removed** mentioned, this is how the guy describe this design:
Very Sorry! I learned about FM detection on Japanese textbook. Therefore I do not know correct English about it. Ratio detector needs a special coil designed to use FM detection. It is not easy to get it in Japan.
The detector here I introduced is called as " Wise detector" . I guess the "Wise" may be the name of the developer. This circuit do not need the special transformer. I can use 455kHz AM coil for it, only changing the frequency by adding the capacitor. By adding 470pF , the frequency of this coil becomes 244kHz. After making this circuit, I connect the signal generator on the (H) point and with changing the frequency, I noted the output voltage of this detector. Please see the right side of the circuit diagram! Here I show you the character of this FM detector.