As LuccasLN notes, there can be many causes for what you describe, and it's hard to suggest causes without more information. If you can, please describe the method in detail, including the injection tables.
If this is a pharmaceutical analysis, then of course it is essential to investigate this phenomenon thoroughly.
THF is a very useful HPLC solvent, but there are some caveats that go along with its use. As you may know, THF is able to dissolve many substances that are insoluble in water and acetonitrile alone, including polymers. It can become contaminated with peroxides that form after it is exposed to oxygen, so it tends to begin absorbing more strongly at short UV wavelengths soon after the bottle is opened; I used to always prepare fresh mobile phases and discard them after a day or two.. Some grades of THF are treated with BHT to prevent peroxide formation; BHT has strong absorption at some UV wavelengths.
Another possible explanation is insufficient elution time, especially with gradient methods. Often, methods are shortened during development to reduce run times, and as a result they may not adequately wash the column. A late eluting peak that remains on the column for several run times can gradually become broader until it is no longer recognizable as a peak. It may appear as a baseline disturbance, which may only affect the later part of the gradient, and may span several runtimes.
In the case of pharmaceutical analysis, a late eluting peak may be a process impurity or degradation product that has not previously been recognized. It also may be an adventitious contaminant. An adventitious contaminant may come from the laboratory environment, the analytical chemist, analytical reagents and solvents, analytical equipment, or the sample. If the source is the sample, then the source may be anywhere in the manufacturing process. It may be something completely unexpected.
