In a peel test, energy is applied to the system in order to separate the adhesive (and its backing) from the substrate. Ideally, all this energy is applied to the adhesive-substrate interface, so that by measuring the applied energy, you then have a value of the strength of that adhesive bond. That however, is not the case in any real system.
Some of the energy applied to the system is spread to other components of the system and does not go to the adhesive-substrate interface. These other components include the part of the adhesive not at the adhesive-substrate interface and also the backing.
Adhesives in general and pressure-sensitive adhesives in particular are made from soft materials which absorb energy. In a peel test, a good amount of the applied energy will be used to first deform the bulk of the adhesive prior to any energy being applied to the adhesive-substrate interface. This is the reason that thicker layers of adhesive have stronger bonds – the actual bond at the interface is identical, but the extra amount of adhesive adsorbs more applied energy. More energy is needed to break the bond. More energy = stronger bond.
Foam adhesives (whether the adhesive is foamed itself or applied to both sides of a layer of foam) are another example of this. The foam absorbs a terrific amount of energy so that the interface doesn't.
A similar issue occurs with the backing – it too must be deformed during the peel test and this deformation requires energy. The stiffer the backing, the more energy that must be applied to deform it.
Both of these deformations then represent energy sinks that make it more difficult to understand the results of any peel test. If a consistent backing is used, and if the adhesive thickness is the same, then it is possible to make comparisons between different adhesives. The energy sinks still exist, but they have become standardized.
Additionally, it is possible with very stiff substrates that the energy sinks will consume such a large fraction of the applied energy (90% +) that the energy applied to breaking the adhesive-substrate interface will be lost in the noise of the measurements. That's when you need to find another test.