Thermostable group II intron reverse transcriptases (TGIRTs) with high fidelity and processivity have been used for a variety of RNA sequencing (RNA-seq) applications, including comprehensive profiling of whole-cell, exosomal, and human plasma RNAs, quantitative tRNA-seq based on the ability of TGIRT enzymes to give full-length reads of tRNAs and other structured small ncRNAs, high-throughput mapping of post-transcriptional modifications, and RNA structure mapping. Here, we improved TGIRT-seq methods for comprehensive transcriptome profiling by (i) rationally designing RNA-seq adapters that minimize adapter dimer formation, and (ii) developing biochemical and computational methods that ameliorate 5’- and 3’-end biases. These improvements, some of which may be applicable to other RNA-seq methods, increase the efficiency of TGIRT-seq library construction and improve the coverage and quantitation of very small RNAs, such as miRNAs. Our findings provide insight into the biochemical basis of 5’- and 3’-end biases in TGIRT-seq and suggest general approaches for ameliorating such biases.