ISO 8601:2004 defines a time interval format which can be expressed in various formats, including "c) by a start and a duration" and "d) by a duration and an end".
Going by the Wikipedia article alone, it seems like the examples only provide for start and end to be "time points", e.g. 2007-03-01T13:00:00Z/P1Y2M10DT2H30M would be a time interval of 1 year, 2 months, 10 days, 2 hours and 3 minutes starting on 2007-03-01 13:00 UTC.
Is it possible to represent a time interval which start or ends after some duration? For example, P1Y/P1Y2M10DT2H30M would be the same duration but "starting" after 1 year.
Such syntax would be useful to model relative time intervals, especially when combined with repeating qualifies. For example, a monthly retirement payout could reasonably be expressed as R/P65Y/1M.
The expression "P1Y/P1Y2M10DT2H30M" is not defined in ISO-8601. This paper mentions following four variants of a "time interval":
a) Start and end are defined as points in time, example => 2019-08-27/2019-08-29
b) Duration without any fixed anchor on the (date)- or timeline, example => P3D
c) Start as point in time and a duration, example => 2019-08-27/P3D
d) A duration and the end as point in time, example => P3D/2019-08-29
So your question "Is it possible to represent a time interval which start or ends after some duration?" can be answered by a clear "No". And honestly said, a double duration expression will confuse most users.