That is a tough one!  I have several ideas in an FCE setting, but with the required consistency for defensible post offer scenereos, that is a tough one!  Two things come to mind in the restriant scenereo-strength and agility.  A formal lifting criteria sounds like a good idea, but how do you quantify this accurately?  At what level(s)?  Certainly hand strength would be a sound criteria, but then again, how do you quantify that?  Reproducability would be the key factor when thinking about using a live person-which is the way I woud go for an FCE, but very hard to construct on a daily/weekly/monthly format.  And how do you do it safely?  You could start with the strength demands for this and similar jobs in the DOT, that might lend us some idea-I'll wager most if not all are in the Medium PDC.  It would be nice to view one of these instances in person, but that sounds like waiting on a corner for a car crash.  One thing is certain-if we-the experts in this field have a hard time quantifying it, detractors surely will!  Here are some ideas;

*See if the facility/company has a training day/refresher course for restraint situations.  Go and view how they teach/train for such instances.  Look for metrics we can use-reproduce grip strenghts uising a dynamometer after wrist holds.  The more samples you have, the better.  Use an average rather than the peak.  Think about pushing/pulling values as well.  Using a sled vs. a force gauge is more reliable, in my opinion.  Think about taking one with you and having lots of weights available and use your force gauge on the sled to document what "feels" accurate for the incumbents after they practice.

*Document your rationale and thinking in a "diary".  Record all values and your math.  Be as open and transparent as possible. 

*Maybe incorporate a mid-level lift as part of your "incumbent" testing?  Maybe a short distance carry from the top of one box to another?  Get feedback from (and document) those that you use as "incumbents" on the day of your analysis.  Using averages vs. peaks again seems smarter.  Think about safety over utility as well.  I recently saw a client for an FCE that was injured during a training/refresher course for restraining disturbed individuals!  She had a wrist injury when a co-worker got carried away in the moment.

Let me know what you think!