As a single DNA molecule is positively supercoiled under constant tension, its extension initially increases due to a negative twist–stretch coupling. The subsequent attainment of an extension maximum has previously been assumed to be indicative of the onset of a phase transition from B- to scP-DNA. Here we show that an extension maximum in fact does not coincide with the onset of a phase transition. This transition is evidenced by a direct observation of a torque plateau using an angular optical trap. Instead we find that the shape of the extension curve can be well explained with a theory that incorporates both DNA twist–stretch coupling and bending fluctuations. This theory also provides a more accurate method of determining the value of the twist–stretch coupling modulus, which has possibly been underestimated in previous studies that did not take into consideration the bending fluctuations. Our study demonstrates the importance of torque detection in the correct identification of phase transitions as well as the contribution of the twist–stretch coupling and bending fluctuations to DNA extension.