You can tell if its just injector noise, or something else, by injecting a matrix blank. That is, try to inject, as exactly as possible, what the sample is dissolved in, without the sample dissolved in it. You can also try repeated injections, to see if this is reproducible. If you keep seeing this negative peak in a sample you're testing for impurities, and it comes out so late its likely not an injector artifact, then you may very well have to quantitate this negative peak -- or more likely, adjust your parameters to try to see its absorbance. Since you're using a PDA @ 254 nm, you can have other wavelength data at that point, if you set up your system to store that data.