I suspect a problem with your schematic. For example, the
NPR transistor that drives the off hook relay would be a poor
design. Most use a PNP transistor that connects to +5V - not
to ground. That would also ground the opto coupler circuit
since the optocoupler should not affect off-hook relay.

The typical destructive transient comes from AC mains
seeking earth ground via the telephone line surge protection.
A most common path to earth is through that PNP transistor,
from relay coil to relay wiper, then on into phone line.

If the transient is small, PNP transistor becomes shorted
causing dial tone to time out when computer power is on. If
the transient is larger, PNP transistors becomes open
circuited resulting in the 'No Dialtone Detected' error
message.

Why a PNP transistor? Higher breakdown voltage between
collector and base provides protection for the modem's
computer. Your schematic provides no such protection - base
to emitter.

If I read everything correctly (and I have not been careful
with the details), it is a classic 'PNP failed shorted'
condition. A 'no dialtone' until computer power is removed so
that relay can open.

wrote:
checked relay-desoldered clicks just fine but is it possible that when
defective it closes
the output without stimulus at the coil?
here is daa schematic
http://www.geocities.com/anglomont/daa.jpg
Since mov is ok I do not see what else can arrest the line...