The Polarized Atomic Hydrogen Jet Target (HJET) Polarimeter commissioned in 2004 was designed to measure absolute polarization of proton beams in Relativistic Heavy Ion Collider (RHIC). HJET performance in RHIC Run 2017 (255 GeV proton beams) will be discussed. The long term (1-100 days) stability of the measured effective analyzing power $\langle A_N\rangle$ was found to be $\sqrt{\langle(\delta\sigma^\mathrm{syst}_{A_N})^2\rangle}/A_N\le0.1\%$. The systematic error in measurement of the RHIC beam average polarization of $P\sim56\%$ was evaluated to be $\sigma_P^\mathrm{syst}/P\lesssim0.5\%$. The statistical error in such a measurement is defined by the number of detected $p^\uparrow p^\uparrow$ elastic events and for an 8 hours RHIC store it was typically at $\sigma_P^\mathrm{stat}\approx2\%$. The elastic events statistics of about $10^9$ per RHIC beam accumulated during Run 2017 allowed us to make a detailed study of the proton-proton single-spin $A_N (t)$ and double-spin $A_{NN}(t)$ analyzing powers in the momentum transfer range $0.001<-t<0.020~(\mathrm{GeV}⁄c)^2$. Contributions of the hadronic single $r_5$ and double $r_2$ spin-flip amplitudes were isolated. Analysis of the inelastic background $p_\mathrm{beam}^\uparrow+p_\mathrm{target}^\uparrow\rightarrow X+p_\mathrm{target}$ allowed us to experimentally evaluate the beam and target spin correlated asymmetries in these processes for $M_X-m_p\lesssim400\ \mathrm{MeV}$ and $t\sim-0.005\ (\mathrm{GeV}⁄c)^2$. The data acquired during 3 weeks of Gold-Gold beams in RHIC allowed us to make a high statistics measurement of the $p^\uparrow\mathrm{Au}$ analyzing power at $E_\text{Lab}=27.2$\unit{}{\GeV}.