The US Federal Communications Commission (FCC) has mandated wireless network operators and mobile devices to provide accurate location information for E-911. Requirements for time of arrival (TOA) and time difference of arrival (TDOA) measurements have been specified in 3GPP LTE Rel. 9 to ensure accurate user equipment (UE) positioning even under bad conditions (e.g. with channel quickly varying and SNR being as low as −13 dB). To fulfil these requirements, it is vital to accurately estimate the first signal arriving path. In this work, we first derive - without any approximation - the Cramér–Rao lower bound (CRLB) of the LTE TOA and TDOA measurements based on the different pilots, which is shown to be as low as a few metres for SNR = −13 dB. The achievable performance of the LTE system is compared with the FCC and 3GPP requirements, and the impact of mobile multipath channels on the measurements is analysed. Then, we describe practical low-complexity methods for LTE TOA and TDOA measurements with enhanced first arriving path detection. The maximum likelihood based correlation profile is used as detection metric. After grossly determining the signal region by a moving window, three methods, namely, peak detection, SNR-based threshold and adaptive threshold based on noise floor and metric peak value are employed to estimate the first arriving path. Simulation results show that the proposed adaptive threshold-based method can meet all 3GPP requirements under various realistic mobile channels, and can in some cases achieve a performance close to the CRLB.