On/offs have the tendency to have a lot of throw. There seem to be a lot of variables like probe response and materials ambient temps. What I have found though lots of testing (hydrofarm and johnson) in cool room temps this is made worse and with thin wall materials worse again (tubs) especially when the heat is attached to the tub directly (peel and stick) Usually the problem is over heating as the probe is often slower to respond than the tub is to heat and the temp shoots far past the set point before the unit shuts power down.

I have an ultratherm and understand what you are saying they are designed to heat slowly and not reach as hot as a regular pad. The thing that would be a problem is if the ultratherm cools faster then the probe. In that case the temp would drop below the set point and enclosure temps with it. I don't know if this would happen or not I don't believe that the pad would drop faster or slower than the probe so I'd guess they would read close to the same so you will only loose the 1-4ºF depending on the unit and when the kick on point is in relationship to the set point.

If you do find a problem you might try (if the ultratherm has enough power to manage it) attaching the pad to glass this will heat and transfer to the enclosure (unless you are using a tank then it is already done) but as glass cools much more slowly the loss will be minimalized. I works on the top end so I figure it will work on the low end too.

On/off units do not have the accuracy or proportional units so there is often probe error tossed into the mix as well. In my experience I have had the same number of on off failures as proportional ones (one each) the on off failed open full power and the proportional safety circuit shut off the unit so it went cold. I actually believe that proportional units although more complex may have a slightly higher failure rate but also have a lower rate of over heating due to the safety systems built into most units. In either case a failsafe is IMO not an optional tool but a necessary one. It prevents damage from a failure. The most common cause of failure of any T-stat is probe being pulled away from the heat usually that is by human error. No single T-stat is safe with out a failsafe in this case.