Since XRS array (~3 arcmin x 3 arcmin) has a comparable size with the XRT point spread function (HPD~2 arcmin), XRS shows a rather complicated response (throughput) to the extended sources. This is demonstrated in the following figure, which is created with ray-tracing simulation using xrssim.
In this simulation, we assumed a point source and uniformly extended sources with several different radii. For a point source (zero radius in the figure), we input 100,000 photons into the telescope, and calculated the number of photons that fall on the XRS pixels (excluding Pixel 2 and 3). For extended sources, surface brightness and number of input photons are adjusted so that the total number of photons included in the XRS Field of view (approximated as a 3 arcmin square) becomes always 100,000. Consequently, as indicated in the top panel, surface brightness of the extended sources is constant for the radii greater than 2.1 arcmin, while increases with smaller radii. In the middle panel, we show the total source photons relative to the case of point source. Total number of photons is 100,000 for the radii smaller than 1.5 arcmin, and it increases in proportional to square of radius for the radii greater than 1.5 arcmin
In the bottom panel, we show the number of photons that fall on the XRS pixels (relative to the case of point source). It is remarkable that the number of photons decreases with the source size and reaches a minimum at about the dimension of XRS. This is because a part of the source photons in the XRS FOV escapes due to XRT point spread function. Then,as the source size exceeds the XRS size, number of photons increases, since those outside photons leak into the XRS FOV. When the source size becomes large enough (>~6 arcmin) so that the effect of PSF tail may be ignored, number of photons on XRS reaches constant.
To summarize, when extended source observation is planned, you will have to take into account the effect of losing photons in the XRS FOV and/or gaining photons from outside of the XRS FOV. Practically, when carrying out spectral simulation for extended sources with XSPEC, please use the released point source ARF, and use the model normalization corresponding to the photon flux from 3 arcmin square area (so is the case for using PIMMS). Then actual number of XRS counts you get will be within +/- 25 % of what XSPEC or PIMMS gives, depending on the source extension. Considering your source size, please adjust the counting rates and/or exposure time. If you would need more precise numbers, please run xrssim using actual sky image of your source.
Finally, it should be reminded that the extended source analysis using ARF is correct only when spectral variation is negligible . If photons in the XRS FOV and those outside have different energy spectra (e.g., cluster of galaxies or SNR having significant spectral variation), more sophisticated spatial-spectral analysis will be required.