Laser tomography and multiconjugate adaptive optics are under development for ground-based extremely large telescopes. Continuous wave sodium guide star lasers are planned for these systems, but their use raises some difficulties due to the extended nature of the beacons generated in the mesosphere and their spatiotemporal variability. We describe a performance analysis on the impact of laser launch telescope (LLT) location for the Thirty Meter Telescope (TMT) multiconjugate adaptive optics system. A semianalytical first-order noise propagation calculation is presented, supplemented by end-to-end Monte Carlo physical optics simulations. The principal conclusion of the study is that modestly superior performance is achieved with multiple LLT locations around the primary mirror, compared to a single central LLT behind the secondary mirror, but the largest value of any of these improvements is of the order of 20 nm rms for the expected wavefront sensor noise levels, suggesting that the final choice of geometry should depend primarily on the cost and complexity of implementation trade-off. This conclusion is also fully supported by the fact that, for the TMT 70 arcsec laser guide star (LGS) asterism, the fratricide effect reduces the performance of the central launch geometry by only a small amount. The reduction ranges from only a few nm rms at zenith to a few tens of nm at a 45°zenith angle in the worst case that the effect cannot be calibrated.