Maybe more of an Econometrics question, but I was wondering why the standard errors between qreg2 and ivqte differ so much (and also the point estimates in the conditional specification), and which one you prefer? Example below that reproduces my case having a binary covariate of interest and a continuous outcome.

Code:
sysuse auto.dta

// unconditional
qreg2 price foreign
ivqte price (foreign), quantiles(.5) variance

// conditional
qreg2 price weight foreign
ivqte price weight (foreign), quantiles(.5) variance