The error report should say which nodes, and you could see
what in the netlist has that connection.
If all pins have DC paths to ground as far as you can see,
check whether symbols may have properties that assign
a connectivity to some internal node that has no pin at
the symbol's schematic level (this is often found in things
like transistors that are generally assumed to have a global
body connection, but might want a more sophisticated
connection model in specific cases, in ID design tools).
Maybe you need to give the property a value like gnd!
to get (say) a global ground plane feature "hooked up"
to the top level.
That's my schematic-free guess (but the netlist and
error report would probably lead you to where you need
to go, if you made the effort). Grep the netlist for the
complained-about node and see how many connections,
and to what, turn up. See also if you can print a node
connections summary and see if things you think are
connected, are really connected. Maybe there's an open
in your subcircuit that you think hits a pin, but doesn't
(or a net is not "implicit" property so doesn't connect
by name the way you think, necessary to assert in some
tools, I do not use this one).
Be sure you're not looking in the wrong place, like the
floating node is really "X1.blahblah" and you're looking at
"blahblah" at the top level out of coincidental naming.
Don't trust the schematic and what you think you see. The
netlist (as boiled down - not necessarily the intermediate
one you get from schematic netlisting, but expanded with
subcircuits' text inlined) is where the floating node beefs
come from. Look there for your clues.