PCL is a flexible polymer at room temperature, hence a decent gripping pressure level is required for the specimens to avoid slippage in the grips. High gripping pressures, however, will induce a high internal stress concentration located in the grips, and failures are expected in that area rather than in the gauge section. The dogbone geometry prevents all those issues, particularly for very ductile or elastomeric materials.
I have found it difficult to use rectangular specimens as they tend to break near the grip. The purpose of the Dog-bone specimen is to help direct stress/strain to the narrowest part of the specimen (the gauge length). It is best to take data from samples which break in the middle. If your samples break in the middle consistently it is an indicator that it could be a suitable sample shape. It would be useful to try both to a suitable standard and compare results.
For international comparison of your polymer, we all should use the same system, otherwise impossible in our commercial work and publication to assess how strong or viscoelastic property the sample has. The system recommended by ISO is standard 37:2011 (Rubber, vulcanized or thermoplastic - Determination of tensile stress-strain properties). Which sample is cut in dog bone or dumb-bell shaped, The size type 3: 20 mm shaft length, 4 mm width. The polymer thickness is usually measured using a digital micrometre.
Alexander Seifalian In my opinion, if your 3D printed samples are made to any fixed size and shape, you can test their mechanical properties for research.