Has anyone experienced the following problem with HP1100 ?

The return to the baseline (UV trace) after a steep gradient (60% in 2 min) becomes longer for no reason and obliges me to increase the time between injections.

You don't give much detail. I presume you're seeing a sloping baseline during that gradient, and the "return" you speak of is the recovery back to the initial baseline level?
A number of things may cause the sloping baseline in the first place. RI effects can be one. Column bleed (not common these days) could be another. However, I suspect you might be seeing components from your sample that "don't want" to elute even under the conditions of the highest %B. Does this make sense in light of your sample?
If so, then you'll have to modify your gradient to ensure that these components do elute with each injection.
If the actual quantities of late eluters is very small (i.e., small amounts of highly "colored" materials), then it may be more practical to run in an isocratic mode and use a guard colum to collect the late eluters. Discard the guard column before break-through occurs.