I see no error
Inputs @ 48 / 96 / 192 kHz
------------------------------------
Mic/Line/Instr front: 4 / 4 / 4
Analog Inputs rear: 8 / 8 / 8
2 x ADAT: 16 / 8 / 4
------------------------------------
TOTAL: 28 / 20 / 16
====================

The total number of inputs and outputs is always on the upper right of product page and is in this case 30.

The further product description is divided into 2 main parts:
- Analog Ports
- Digital Ports
In these chapters the I/O capabilities of each type of port are detailed in regards to
- typical use cases and
- number of ports which you achieve depending on sample rate

Exmple:
1. for SPDIF it tells you that you will have 2 channels no matter which sample rate you choose
2. ADAT can be taken to i.e. expand the unit for AD/DA and makes the calculation, how many channels you get depending on the sample rate. So you get 16 by ADAT and 4 by the front inputs. And in this case it details how you need to count with higher sample rates as not everybody knows about ADAT multiplexing for higher samplerates.

You need to see the whole picture how this product page is structured, not only the chapter.

Just trying to help clear up what looked goofy when I read it, I almost didn't make this post.

However, I think you just touched on a key misunderstanding:
"Adding AES to this setup makes no real sense as no one would use such a setup"

I have the very thing here to get a total of 30 192kHz channels using two ADAT and two AES cables feeding an 800 and an 802.  In my case it exactly maximizes channel use of my hardware.  Why would somebody ignore available input on an interface when it will work just fine ?

Which was sort of my original point.  The AES/EBU sentence doesn't mention the 12 analogs.  The ADAT section then turns around and counts the analog inputs while excluding AES and then describing the unit as a whole with "the 802 will provide".

It just struck me as an inconsistent to someone learning about the device to alternately leave out the analog inputs in one sentence, then count then in the next while making it sound like that's the maximum capability of the FF802.

If I didn't know better and were still shopping for an interface, I'd be trying to figure out why the count didn't seem to match the table in the Inputs/Outputs description.

Maybe I'm just sensitive to inconsistencies because I've been going through hell with all the glitches in TI's C5000 DSP ecosystem, as well as having to read through Lauterbach's TRACE32 documents because the website product description is vague in places.