Based on observations obtained at the New Technology Telescope, La Silla, Chile.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

We also assessed the HeII $\lambda$ 5411 line, but discarded the measurements because of the relatively low S/N ratio in $\sim$ 50% of the spectra.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... RVs

The situation is slightly more complicated since HeII $\lambda$ 4025 is in fact a blend of the lines HeII $\lambda$ 4025.60 and HeI $\lambda$ 4026.191. Moreover, the Balmer lines H $\delta$ and H $\gamma$ are blended with HeII lines as well. In order to be perfectly precise, the positions of the profiles measured on the spectra should be compared to the rest wavelength of the corresponding lines. But in the case of blends, the contribution of each individual line is not known. We make here the reasonable hypothesis that the relative intensity contribution of one individual line to a given blend is not changing with time. Therefore, the comparison of a profile position to one rest wavelength is relevant and coherent. Thus the choice of one rest wavelength to another will only produce an offset in the absolute value of the RVs. At that point of our analysis, only relative RVs are meaningful. In general, except when specifically stated, the rest wavelength that is used is the one of the line specified explicitly in the text.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... lines

For the H+He blends, the simple average of the rest wavelength has been taken. The important point here is to compare the velocities of the absorption and emission profiles together (see also previous footnote).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... value

We were not able to compute the correct MACHO magnitude, following Eqs. (1) and (2) of Alcock et al. (1999), since some coefficient values were not accessible on the Internet, at the time of writing.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.