Signal information is lost by using Decimator block!!

I am building a model regarding fmcw radar whose output is 600000 samples for one sweep. After dechirping at receivers side , both targets are detected but when I downsample my information to 2000 samples per sweep my second target is not being detected accurately ( i.e it's range is not calculated as desired) Right after dechirp block, peaks of both targets are absolutely perfect but after decimation, the second peak is lost somehow. Any help?

My fft length is 2048 , that is why I downsampled it to 2000 samples. Is there any alternative to preserve the data? I cannot use more thn 2048 length