In a theoretical world with a balanced load and lossless transmission line, you are correct. However, in a real world application other factors apply. In the 22nd Edition of the ARRL Antenna book (circa 2011) in section 23.2.6 Input Impedance the effect of line loss on input impedance is calculated, which agrees with the Transmission Line Program for Windows program V 3.24 by N6BV supplied with the Antenna Book. A load of 40+j30 with 50' of RG-213 has an input Z of 65.8+j32 and the SWR at the load of 1.84:1 is increased to 1.96:1 at the input. This can be confirmed with an AIM 4170C analyzer (using properly constructed common-mode chokes) to reduce the inherent unbalanced current in a real dipole).
Many SWR meters are sensitive to current flowing on the outside of the shield and that is indeed the more typical cause of changing SWR with line length, which can be reduced with a CM choke at the feed point of the antenna.
The effects of line length and loss are much more dramatic at VHF and above. Having spent hours fine tuning transmission lines between duplexer cavities and then between the duplexer TX output and the repeater. Losses are much greater as frequencies increase.
For HF bands line length would not account for the 2:1 to 3:1 variation the original poster was reporting, but it can be significant above 20 MHz, especially with solid dielectric coax.