Yes, but wouldn't it have to, in order to detect a loop?
Reading
https://gtacknowledge.extremenetworks.com/articles/Q_A/What-is-ELRP, it says that the switch sends out a packet on every configured (including the uplink, in my approach) port, and sees if it gets it back on another port.
If I have a switch scenario like:
Core-Switch
/ \
EdgeSW-1 EdgeSW-2and have a vlan with untagged port 1 on Edge-1 and Edge-2, and tagged ports 55 on Edge-1 and Edge-2 (the uplink ports), same vlan on the Core between Edge-1 and Edge-2, and configure elrp both on Edge-1 and Edge-2, wouldn't the only way that I can detect a loop between Edge-1 and Edge-2 (let's just say 'physical cable') be because I the packet goes out port-1 Edge-1 and comes in on Edge-1 port 55 - or out Edge-1 55 and back in on Edge-1 port 1?
If I wouldn't include the uplink port on the elrp port list (OK, 'all vlan ports'), I'd never detect the loop, even if it existed.
Now, if you're only concerned with loops on a single Edge switch (for instance because you don't have the same vlan on another Edge switch, then you could live without the uplink port in ELRP - but I fear you'd then have to specify every port that's in the vlan, except the uplink port, and only ports that are part of that vlan.
And if you add ports to the vlan, you'd probably have to add them manually to the ELRP list. And that would just drive me nuts, because I know I'd screw up sooner or later (but that's juist me )
Standard disclaimer - I may have something wrong. Also, I only have BlackDiamon/Summit 460/480/670s running EXOS 15.4 or higher.