How to get rid of NaN result when using setEvidence function in gRain package when doing Bayesian modeling?

For context, I am conducting hazard modelling, examining the flow of impacts from node to node within a network using Bayesian network analysis. I have a simplified network diagram dag below which I will need to scale up for a more complex network later.

I want to create a hypothetical scenario where node A1 is in an "Interrupted" state. I am using setEvidence function, but when I run a query of the state of node A3, I am getting "NaN" as the result.

Any idea on how to resolve this? Below is the simplified code.

# Load required libraries
library(bnlearn)
library(gRain)

# Define states
states <- c("Normal", "Interrupted")

# Define CPTs
cptA <- matrix(c(1, 0),
               nrow = 1,
               dimnames = list(NULL, states))

cptA1 <- array(
  c(1, 0,
    0, 1),
  dim = c(2, 2),
  dimnames = list(
    A1 = states,
    A = states
  )
)

cptA2 <- array(
  c(1, 0,
    0, 1),
  dim = c(2, 2),
  dimnames = list(
    A2 = states,
    A1 = states
  )
)

cptA3 <- array(
  c(1, 0,
    0, 1),
  dim = c(2, 2),
  dimnames = list(
    A3 = states,
    A2 = states
  )
)

# Build DAG
dag <- model2network("[A][A1|A][A2|A1][A3|A2]")

# Fit the Bayesian network
fitted <- custom.fit(dag, dist = list(
  A = cptA,
  A1 = cptA1,
  A2 = cptA2,
  A3 = cptA3
))

# Convert to gRain object for inference
grain_net <- as.grain(fitted)

# Set evidence: force A1 = "Interrupted"
grain_evidence <- setEvidence(grain_net, evidence = list(A1 = "Interrupted"))

# Query the posterior of A3
query_result <- querygrain(grain_evidence, nodes = "A3")

# Print the result
print(query_result)