That’s true. Erlang B assumes that all blocked calls are immediately cleared from the system.
Which model you use depends on how long the blocked calls remain in the system. If they remain in the system only as long as their holding time (had they been answered), then you should use the Poisson (Molina) model. If you are only looking at 1% blocking, the error introduced by using Erlang B would be almost negligible.
If calls are held in the system until they are answered, then Erlang C is the correct model to use for analysing this form of queuing.
We do not have products that support Poisson (Molina), because this situation is not often found in the real world (blocked calls are usually cleared or queued). We do have Erlang C Calculators though.
You may also wish to consider the Extended Erlang B model. There is a free calculator on this site. It is used to model the situation where a percentage of blocked calls immediately retry. Again, for 1% blocking, you probably wouldn’t notice a difference in the results anyway.